You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
1153 lines
52 KiB
1153 lines
52 KiB
#!/usr/bin/perl |
|
# |
|
# This scans the cluster by processing the CIB and storing information about nodes, cluster states, etc. |
|
# |
|
# NOTE: The data stored here is not bound to a given host. As such, node 1 will run this agent and node 2 |
|
# won't, _except_ when node 1 is offline and only node 2 is up. |
|
# |
|
# Examples; |
|
# |
|
# Exit codes; |
|
# 0 = Normal exit. |
|
# 1 = Startup failure (not running as root, no DB, bad file read, etc) |
|
# 2 = Not a cluster member |
|
# |
|
# TODO: |
|
# - When a node is lost, update the location constraints to keep the servers on the surviving node when the |
|
# peer returns. |
|
# |
|
|
|
use strict; |
|
use warnings; |
|
use Anvil::Tools; |
|
use Data::Dumper; |
|
use Text::Diff; |
|
|
|
# Disable buffering |
|
$| = 1; |
|
|
|
# Prevent a discrepency between UID/GID and EUID/EGID from throwing an error. |
|
$< = $>; |
|
$( = $); |
|
|
|
my $THIS_FILE = ($0 =~ /^.*\/(.*)$/)[0]; |
|
my $running_directory = ($0 =~ /^(.*?)\/$THIS_FILE$/)[0]; |
|
if (($running_directory =~ /^\./) && ($ENV{PWD})) |
|
{ |
|
$running_directory =~ s/^\./$ENV{PWD}/; |
|
} |
|
|
|
my $anvil = Anvil::Tools->new(); |
|
|
|
# Make sure we're running as 'root' |
|
# $< == real UID, $> == effective UID |
|
if (($< != 0) && ($> != 0)) |
|
{ |
|
# Not root |
|
print $anvil->Words->string({key => "error_0005"})."\n"; |
|
$anvil->nice_exit({exit_code => 1}); |
|
} |
|
|
|
$anvil->data->{scancore}{'scan-cluster'}{disable} = 0; |
|
$anvil->data->{switches}{force} = 0; |
|
|
|
$anvil->Storage->read_config(); |
|
|
|
# Read switches |
|
$anvil->Get->switches; |
|
|
|
# Handle start-up tasks |
|
my $problem = $anvil->ScanCore->agent_startup({agent => $THIS_FILE}); |
|
if ($problem) |
|
{ |
|
$anvil->nice_exit({exit_code => 1}); |
|
} |
|
|
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 2, key => "log_0115", variables => { program => $THIS_FILE }}); |
|
if ($anvil->data->{switches}{purge}) |
|
{ |
|
# This can be called when doing bulk-database purges. |
|
my $schema_file = $anvil->data->{path}{directories}{scan_agents}."/".$THIS_FILE."/".$THIS_FILE.".sql"; |
|
$anvil->Database->purge_data({ |
|
debug => 2, |
|
tables => $anvil->Database->get_tables_from_schema({schema_file => $schema_file}), |
|
}); |
|
$anvil->nice_exit({exit_code => 0}); |
|
} |
|
|
|
# Before we do anything, are we a node in a pacemaker cluster? |
|
my $host_type = $anvil->Get->host_type; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => { host_type => $host_type }}); |
|
if ($host_type ne "node") |
|
{ |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 2, key => "scan_cluster_log_0002", variables => { host_type => $host_type }}); |
|
$anvil->nice_exit({exit_code => 0}); |
|
} |
|
|
|
# Read last scan |
|
read_last_scan($anvil); |
|
|
|
# Read and process in one shot. |
|
collect_data($anvil); |
|
|
|
# Find changes. |
|
find_changes($anvil); |
|
|
|
# Check the cluster config. |
|
check_config($anvil); |
|
|
|
# Check the fence delay |
|
check_fence_delay($anvil); |
|
|
|
# Check for failed resources or resources that need updates |
|
check_resources($anvil); |
|
|
|
# Check for stale data in the CIB and clean up any that are found |
|
cib_cleanup($anvil); |
|
|
|
# Shut down. |
|
$anvil->ScanCore->agent_shutdown({agent => $THIS_FILE}); |
|
|
|
|
|
############################################################################################################# |
|
# Functions # |
|
############################################################################################################# |
|
|
|
|
|
sub cib_cleanup |
|
{ |
|
my ($anvil) = @_; |
|
|
|
my $problem = $anvil->Cluster->parse_cib({debug => 2}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { problem => $problem }}); |
|
if ($problem) |
|
{ |
|
# Not in a cluster |
|
return(0); |
|
} |
|
|
|
# Find the servers still on the cluster |
|
foreach my $server_name (sort {$a cmp $b} keys %{$anvil->data->{cib}{parsed}{data}{server}}) |
|
{ |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { server_name => $server_name }}); |
|
if (exists $anvil->data->{cib}{parsed}{data}{server}{$server_name}{drbd_fence_rule}{'exists'}) |
|
{ |
|
my $attribute = $anvil->data->{cib}{parsed}{data}{server}{$server_name}{drbd_fence_rule}{attribute}; |
|
my $operation = $anvil->data->{cib}{parsed}{data}{server}{$server_name}{drbd_fence_rule}{operation}; |
|
my $value = $anvil->data->{cib}{parsed}{data}{server}{$server_name}{drbd_fence_rule}{value}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
server_name => $server_name, |
|
attribute => $attribute, |
|
operation => $operation, |
|
value => $value, |
|
}}); |
|
} |
|
} |
|
|
|
my $host_name = $anvil->Get->host_name; |
|
my $short_host_name = $anvil->Get->short_host_name; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
host_name => $host_name, |
|
short_host_name => $short_host_name, |
|
}}); |
|
|
|
my $node_id = ""; |
|
my $node_name = ""; |
|
foreach my $node (sort {$a cmp $b} keys %{$anvil->data->{cib}{parsed}{data}{node}}) |
|
{ |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { node => $node }}); |
|
if (($node eq $short_host_name) or ($node eq $host_name)) |
|
{ |
|
$node_id = $anvil->data->{cib}{parsed}{data}{node}{$node}{id}; |
|
$node_name = $anvil->data->{cib}{parsed}{configuration}{nodes}{$node_id}{uname}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
node_id => $node_id, |
|
node_name => $node_name, |
|
}}); |
|
last; |
|
} |
|
} |
|
if ($node_id eq "") |
|
{ |
|
# Node ID for this node was not found! |
|
return(0); |
|
} |
|
|
|
foreach my $attribute_id (sort {$a cmp $b} keys %{$anvil->data->{cib}{parsed}{cib}{node_state}{$node_id}}) |
|
{ |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { attribute_id => $attribute_id }}); |
|
if ($attribute_id =~ /^drbd-fenced_(.*)$/) |
|
{ |
|
my $server_name = $1; |
|
my $state = $anvil->data->{cib}{parsed}{cib}{node_state}{$node_id}{$attribute_id}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
server_name => $server_name, |
|
'state' => $state, |
|
}}); |
|
|
|
if (not exists $anvil->data->{cib}{parsed}{data}{server}{$server_name}) |
|
{ |
|
# Stale attribute, remove it! |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0012", variables => { attribute => $attribute_id }}); |
|
|
|
my $shell_call = $anvil->data->{path}{exe}{pcs}." node attribute ".$node_name." ".$attribute_id."="; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { shell_call => $shell_call }}); |
|
|
|
my ($output, $return_code) = $anvil->System->call({shell_call => $shell_call}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
output => $output, |
|
return_code => $return_code, |
|
}}); |
|
} |
|
} |
|
} |
|
|
|
return(0); |
|
} |
|
|
|
### TODO: Rework this to use Server->connect_to_libvirtd and phase out Server->find(). |
|
# This looks for failed resource and, if found, tries to recover them. |
|
sub check_resources |
|
{ |
|
my ($anvil) = @_; |
|
|
|
my ($problem) = $anvil->Cluster->parse_cib({debug => 2}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { problem => $problem }}); |
|
|
|
foreach my $server (sort {$a cmp $b} keys %{$anvil->data->{crm_mon}{parsed}{'pacemaker-result'}{resources}{resource}}) |
|
{ |
|
# This is used for alerts, if needed below. |
|
my $variables = { server => $server }; |
|
|
|
my $failed = check_if_server_failed($anvil, $server); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
server => $server, |
|
failed => $failed, |
|
}}); |
|
if ($failed) |
|
{ |
|
# Who am I and who is my peer? See if the server is running on either host. |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0003", variables => { server => $server }}); |
|
my $attempt_recovery = 0; |
|
my $server_found = 0; |
|
my $both_nodes_ready = 1; |
|
foreach my $target ("local", "peer") |
|
{ |
|
my $node_ready = $anvil->data->{cib}{parsed}{$target}{ready}; |
|
my $node_name = $anvil->data->{cib}{parsed}{$target}{name}; |
|
my $host_uuid = $anvil->Get->host_uuid_from_name({host_name => $node_name}); |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0004", variables => { |
|
node_name => $node_name, |
|
host_uuid => $host_uuid, |
|
node_ready => $node_ready, |
|
}}); |
|
if (not $node_ready) |
|
{ |
|
$both_nodes_ready = 1; |
|
} |
|
|
|
if ($host_uuid eq $anvil->Get->host_uuid) |
|
{ |
|
# Search for the server here |
|
$anvil->Server->find({debug => 2}); |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0005"}); |
|
} |
|
else |
|
{ |
|
# Search for the server on the peer. |
|
my $target_ip = $anvil->Network->find_target_ip({host_uuid => $host_uuid}); |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0006", variables => { target_ip => $target_ip }}); |
|
$anvil->Server->find({ |
|
debug => 2, |
|
target => $target_ip, |
|
}); |
|
} |
|
my $server_host = defined $anvil->data->{server}{location}{$server}{host_name} ? $anvil->data->{server}{location}{$server}{host_name} : ""; |
|
my $server_status = defined $anvil->data->{server}{location}{$server}{status} ? $anvil->data->{server}{location}{$server}{status} : ""; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 1, list => { |
|
server_host => $server_host, |
|
server_status => $server_status, |
|
}}); |
|
if ($server_host) |
|
{ |
|
$server_found = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { server_found => $server_found }}); |
|
if (($node_ready) && ($host_uuid eq $anvil->Get->host_uuid)) |
|
{ |
|
# Go ahead with recovery |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0007"}); |
|
$attempt_recovery = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { attempt_recovery => $attempt_recovery }}); |
|
} |
|
} |
|
} |
|
|
|
if ((not $server_found) && ($both_nodes_ready)) |
|
{ |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0008"}); |
|
$attempt_recovery = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { attempt_recovery => $attempt_recovery }}); |
|
} |
|
elsif (($server_found) && (not $attempt_recovery)) |
|
{ |
|
# The server was found to be running, but not here (or this node is not fully in the cluster). NOT attempting recovery yet. |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0009"}); |
|
} |
|
|
|
if ($attempt_recovery) |
|
{ |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0010"}); |
|
$anvil->Cluster->recover_server({ |
|
debug => 2, |
|
server => $server, |
|
running => $server_found, |
|
}); |
|
|
|
# It'll leave 'failed state' for a bit, so we need to wait. |
|
sleep 3; |
|
my $wait_until = time + 10; |
|
my $waiting = 1; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_log_0011"}); |
|
while($waiting) |
|
{ |
|
my $failed = check_if_server_failed($anvil, $server); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { failed => $failed }}); |
|
if ($failed) |
|
{ |
|
# No luck... |
|
$waiting = 0; |
|
my $changed = $anvil->Alert->check_alert_sent({ |
|
record_locator => "scan_cluster::failed_server::".$server, |
|
set_by => $THIS_FILE, |
|
}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
waiting => $waiting, |
|
changed => $changed, |
|
}}); |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_alert_0014", variables => $variables}); |
|
if ($changed) |
|
{ |
|
# Send an alert. |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0014", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
} |
|
elsif (time > $wait_until) |
|
{ |
|
# Success! |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_alert_0013", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0013", variables => $variables, set_by => $THIS_FILE}); |
|
|
|
# Clear the alert, if it existed before |
|
$waiting = 0; |
|
my $changed = $anvil->Alert->check_alert_sent({ |
|
record_locator => "scan_cluster::failed_server::".$server, |
|
set_by => $THIS_FILE, |
|
clear => 1, |
|
}); |
|
} |
|
else |
|
{ |
|
# Wait a sec |
|
sleep 2; |
|
} |
|
} |
|
} |
|
} |
|
else |
|
{ |
|
# Make sure that this server wasn't previously failed. |
|
my $changed = $anvil->Alert->check_alert_sent({ |
|
record_locator => "scan_cluster::failed_server::".$server, |
|
set_by => $THIS_FILE, |
|
clear => 1, |
|
}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { changed => $changed }}); |
|
if ($changed) |
|
{ |
|
# Send the All-good alert. |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 1, key => "scan_cluster_alert_0015", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0015", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
|
|
# Check to see if server's parameters need updating |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { server => $server }}); |
|
my $update = 0; |
|
foreach my $op (sort {$a cmp $b} keys %{$anvil->data->{cib}{parsed}{cib}{resources}{primitive}{$server}{operations}{op}}) |
|
{ |
|
my $name = $anvil->data->{cib}{parsed}{cib}{resources}{primitive}{$server}{operations}{op}{$op}{name}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { name => $name }}); |
|
if (($name eq "migrate_to") or ($name eq "migrate_from")) |
|
{ |
|
my $on_fail = exists $anvil->data->{cib}{parsed}{cib}{resources}{primitive}{$server}{operations}{op}{$op}{'on-fail'} ? $anvil->data->{cib}{parsed}{cib}{resources}{primitive}{$server}{operations}{op}{$op}{'on-fail'} : ""; |
|
my $timeout = exists $anvil->data->{cib}{parsed}{cib}{resources}{primitive}{$server}{operations}{op}{$op}{timeout} ? $anvil->data->{cib}{parsed}{cib}{resources}{primitive}{$server}{operations}{op}{$op}{timeout} : 0; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
on_fail => $on_fail, |
|
timeout => $timeout, |
|
}}); |
|
if (($on_fail ne "block") or ($timeout != 600)) |
|
{ |
|
$update = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { update => $update }}); |
|
} |
|
} |
|
} |
|
|
|
if ($update) |
|
{ |
|
# Update the resource config. |
|
my $resource_command = $anvil->data->{path}{exe}{pcs}." resource update ".$server." ocf:alteeve:server name=\"".$server."\" meta allow-migrate=\"true\" op monitor interval=\"60\" start timeout=\"60\" on-fail=\"block\" stop timeout=\"300\" on-fail=\"block\" migrate_to timeout=\"600\" on-fail=\"block\" migrate_from timeout=\"600\" on-fail=\"block\""; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "log_0739", variables => { |
|
server => $server, |
|
command => $resource_command, |
|
}}); |
|
|
|
my ($output, $return_code) = $anvil->System->call({shell_call => $resource_command}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
output => $output, |
|
return_code => $return_code, |
|
}}); |
|
} |
|
} |
|
} |
|
|
|
return(0); |
|
} |
|
|
|
sub check_if_server_failed |
|
{ |
|
my ($anvil, $server) = @_; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => { server => $server }}); |
|
|
|
$anvil->Cluster->parse_crm_mon({debug => 2}); |
|
my $failed = exists $anvil->data->{crm_mon}{parsed}{'pacemaker-result'}{resources}{resource}{$server}{variables}{failed} ? $anvil->data->{crm_mon}{parsed}{'pacemaker-result'}{resources}{resource}{$server}{variables}{failed} : 0; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => { failed => $failed }}); |
|
if ($failed eq "true") |
|
{ |
|
$failed = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => { failed => $failed }}); |
|
} |
|
elsif ($failed eq "false") |
|
{ |
|
$failed = 0; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => { failed => $failed }}); |
|
} |
|
|
|
return($failed); |
|
} |
|
|
|
# Check to see if we need to move the fence delay. |
|
sub check_fence_delay |
|
{ |
|
my ($anvil) = @_; |
|
|
|
my $preferred_node = $anvil->Cluster->manage_fence_delay(); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { preferred_node => $preferred_node }}); |
|
if ($preferred_node ne "!!error!!") |
|
{ |
|
### NOTE: We don't make the peer be the preferred node, a node can only make itself the preferred |
|
### node. |
|
# How many servers are running on each node. |
|
$anvil->Database->get_anvils(); |
|
$anvil->Database->get_servers(); |
|
$anvil->Cluster->get_peers(); |
|
my $anvil_uuid = $anvil->Cluster->get_anvil_uuid(); |
|
my $local_node_is = $anvil->data->{sys}{anvil}{i_am}; |
|
my $local_node_name = $anvil->data->{cib}{parsed}{'local'}{name}; |
|
my $local_host_name = $anvil->data->{sys}{anvil}{$local_node_is}{host_name}; |
|
my $local_host_uuid = $anvil->data->{sys}{anvil}{$local_node_is}{host_uuid}; |
|
my $peer_node_is = $anvil->data->{sys}{anvil}{peer_is}; |
|
my $peer_node_name = $anvil->data->{cib}{parsed}{peer}{name};; |
|
my $peer_host_name = $anvil->data->{sys}{anvil}{$peer_node_is}{host_name}; |
|
my $peer_host_uuid = $anvil->data->{sys}{anvil}{$peer_node_is}{host_uuid}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
anvil_uuid => $anvil_uuid, |
|
local_node_is => $local_node_is, |
|
local_node_name => $local_node_name, |
|
local_host_name => $local_host_name, |
|
local_host_uuid => $local_host_uuid, |
|
peer_node_is => $peer_node_is, |
|
peer_node_name => $peer_node_name, |
|
peer_host_name => $peer_host_name, |
|
peer_host_uuid => $peer_host_uuid, |
|
}}); |
|
|
|
# Get the short host names, as that's usually what the node name is. |
|
my $local_short_host_name = $local_host_name; |
|
$local_short_host_name =~ s/\..$//; |
|
my $peer_short_host_name = $peer_host_name; |
|
$peer_short_host_name =~ s/\..$//; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
local_short_host_name => $local_short_host_name, |
|
peer_short_host_name => $peer_short_host_name, |
|
}}); |
|
|
|
# If my peer isn't in the cluster, make sure I am the fence delay host. |
|
if (not $anvil->data->{cib}{parsed}{peer}{ready}) |
|
{ |
|
# My peer is not ready, make sure I'm the preferred host. |
|
if (($preferred_node eq $local_node_name) or ($preferred_node eq $local_host_name) && ($preferred_node eq $local_short_host_name)) |
|
{ |
|
# We're good. |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 2, key => "log_0633"}); |
|
} |
|
else |
|
{ |
|
# We're not, set the delay to us. |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "log_0634"}); |
|
my $preferred_node = $anvil->Cluster->manage_fence_delay({prefer => $local_node_name}); |
|
return(0); |
|
} |
|
} |
|
|
|
# How many servers are on each node? |
|
my $local_server_count = 0; |
|
my $peer_server_count = 0; |
|
foreach my $server_uuid (keys %{$anvil->data->{servers}{server_uuid}}) |
|
{ |
|
next if $anvil_uuid ne $anvil->data->{servers}{server_uuid}{$server_uuid}{server_anvil_uuid}; |
|
|
|
my $server_name = $anvil->data->{servers}{server_uuid}{$server_uuid}{server_name}; |
|
my $server_state = $anvil->data->{servers}{server_uuid}{$server_uuid}{server_state}; |
|
my $server_host_uuid = $anvil->data->{servers}{server_uuid}{$server_uuid}{server_host_uuid}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
server_uuid => $server_uuid, |
|
server_name => $server_name, |
|
server_state => $server_state, |
|
server_host_uuid => $server_host_uuid, |
|
}}); |
|
next if $server_state eq "shut off"; |
|
if ($server_state eq "migrating") |
|
{ |
|
# Don't do anything. |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "log_0635", variables => { server_name => $server_name }}); |
|
return(0); |
|
} |
|
if ($server_host_uuid eq $local_host_uuid) |
|
{ |
|
$local_server_count++; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { local_server_count => $local_server_count }}); |
|
} |
|
elsif ($server_host_uuid eq $peer_host_uuid) |
|
{ |
|
$peer_server_count++; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { peer_server_count => $peer_server_count }}); |
|
} |
|
} |
|
|
|
# Don't do anything if there are no servers running anywhere, or if both servers have at least one |
|
# server. |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
local_server_count => $local_server_count, |
|
peer_server_count => $peer_server_count, |
|
}}); |
|
if ((not $local_server_count) && (not $peer_server_count)) |
|
{ |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 2, key => "log_0636"}); |
|
return(0); |
|
} |
|
elsif (($local_server_count) && ($peer_server_count)) |
|
{ |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 2, key => "log_0637", variables => { |
|
local_server_count => $local_server_count, |
|
peer_server_count => $peer_server_count, |
|
}}); |
|
return(0); |
|
} |
|
elsif (($local_server_count) && ($preferred_node ne $local_node_name)) |
|
{ |
|
# Make us the preferred host. |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "log_0638"}); |
|
my $preferred_node = $anvil->Cluster->manage_fence_delay({ |
|
debug => 2, |
|
prefer => $local_node_name, |
|
}); |
|
return(0); |
|
} |
|
} |
|
|
|
return(0); |
|
} |
|
|
|
sub check_config |
|
{ |
|
my ($anvil) = @_; |
|
|
|
$anvil->Database->get_manifests(); |
|
my $anvil_name = $anvil->Cluster->get_anvil_name({}); |
|
my $manifest_uuid = $anvil->data->{manifests}{manifest_name}{$anvil_name}{manifest_uuid}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
anvil_name => $anvil_name, |
|
manifest_uuid => $manifest_uuid, |
|
}}); |
|
|
|
if ($manifest_uuid) |
|
{ |
|
# Check to see if the stonith config needs to be updated. |
|
$anvil->Cluster->check_stonith_config({debug => 2}); |
|
} |
|
|
|
return(0); |
|
} |
|
|
|
# Looks for changes. |
|
sub find_changes |
|
{ |
|
my ($anvil) = @_; |
|
|
|
# We can't track a cluster through name change, so either we're INSERTing a new one, or bust. |
|
my $scan_cluster_anvil_uuid = $anvil->Cluster->get_anvil_uuid(); |
|
my $anvil_name = $anvil->Get->anvil_name_from_uuid({anvil_uuid => $scan_cluster_anvil_uuid}); |
|
my $scan_cluster_uuid = ""; |
|
my $cluster_name = $anvil->data->{cib}{parsed}{data}{cluster}{name}; |
|
my $stonith_enabled = $anvil->data->{cib}{parsed}{data}{stonith}{enabled}; |
|
my $stonith_max_attempts = $anvil->data->{cib}{parsed}{data}{stonith}{'max-attempts'}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
cluster_name => $cluster_name, |
|
stonith_enabled => $stonith_enabled, |
|
stonith_max_attempts => $stonith_max_attempts, |
|
}}); |
|
|
|
# If we're a full cluster member, read the CIB as well. |
|
my $cluster_cib = ""; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { "cib::parsed::local::ready" => $anvil->data->{cib}{parsed}{'local'}{ready} }}); |
|
if ($anvil->data->{cib}{parsed}{'local'}{ready}) |
|
{ |
|
($cluster_cib, my $return_code) = $anvil->System->call({shell_call => $anvil->data->{path}{exe}{pcs}." cluster cib", source => $THIS_FILE, line => __LINE__}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
cluster_cib => $cluster_cib, |
|
return_code => $return_code, |
|
}}); |
|
} |
|
|
|
if (exists $anvil->data->{sql}{anvil_uuid}{$scan_cluster_anvil_uuid}) |
|
{ |
|
# Check for a name change |
|
$scan_cluster_uuid = $anvil->data->{sql}{anvil_uuid}{$scan_cluster_anvil_uuid}; |
|
my $old_cluster_name = $anvil->data->{sql}{scan_cluster}{scan_cluster_uuid}{$scan_cluster_uuid}{scan_cluster_name}; |
|
my $old_cluster_cib = $anvil->data->{sql}{scan_cluster}{scan_cluster_uuid}{$scan_cluster_uuid}{scan_cluster_cib}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
scan_cluster_uuid => $scan_cluster_uuid, |
|
old_cluster_name => $old_cluster_name, |
|
old_cluster_cib => $old_cluster_cib, |
|
}}); |
|
if ($cluster_name ne $old_cluster_name) |
|
{ |
|
# The name of the cluster has changed. |
|
my $query = " |
|
UPDATE |
|
scan_cluster |
|
SET |
|
scan_cluster_name = ".$anvil->Database->quote($cluster_name).", |
|
modified_date = ".$anvil->Database->quote($anvil->Database->refresh_timestamp)." |
|
WHERE |
|
scan_cluster_uuid = ".$anvil->Database->quote($scan_cluster_uuid)." |
|
;"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
$anvil->Database->write({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
|
|
my $variables = { |
|
new_cluster_name => $cluster_name, |
|
old_cluster_name => $old_cluster_name, |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0002", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0002", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
if (($cluster_cib) && ($cluster_cib ne $old_cluster_cib)) |
|
{ |
|
my $query = " |
|
UPDATE |
|
scan_cluster |
|
SET |
|
scan_cluster_cib = ".$anvil->Database->quote($cluster_cib).", |
|
modified_date = ".$anvil->Database->quote($anvil->Database->refresh_timestamp)." |
|
WHERE |
|
scan_cluster_uuid = ".$anvil->Database->quote($scan_cluster_uuid)." |
|
;"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
$anvil->Database->write({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
|
|
my $difference = diff \$old_cluster_cib, \$cluster_cib, { STYLE => 'Unified' }; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { difference => $difference }}); |
|
|
|
my $variables = { |
|
cluster_name => $cluster_name, |
|
difference => $difference, |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0012", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0012", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
} |
|
else |
|
{ |
|
# New cluster, INSERT |
|
$scan_cluster_uuid = $anvil->Get->uuid(); |
|
my $query = " |
|
INSERT INTO |
|
scan_cluster |
|
( |
|
scan_cluster_uuid, |
|
scan_cluster_anvil_uuid, |
|
scan_cluster_name, |
|
scan_cluster_cib, |
|
modified_date |
|
) VALUES ( |
|
".$anvil->Database->quote($scan_cluster_uuid).", |
|
".$anvil->Database->quote($scan_cluster_anvil_uuid).", |
|
".$anvil->Database->quote($cluster_name).", |
|
".$anvil->Database->quote($cluster_cib).", |
|
".$anvil->Database->quote($anvil->Database->refresh_timestamp)." |
|
);"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
$anvil->Database->write({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
|
|
my $variables = { cluster_name => $cluster_name }; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0001", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0001", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
|
|
$anvil->Database->get_anvils(); |
|
foreach my $scan_cluster_node_name (sort {$a cmp $b} keys %{$anvil->data->{cib}{parsed}{data}{node}}) |
|
{ |
|
my $scan_cluster_node_host_uuid = $anvil->Get->host_uuid_from_name({host_name => $scan_cluster_node_name}) // ""; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
scan_cluster_node_name => $scan_cluster_node_name, |
|
scan_cluster_node_host_uuid => $scan_cluster_node_host_uuid, |
|
}}); |
|
if (not $scan_cluster_node_host_uuid) |
|
{ |
|
# Something is wrong with this host. Does the hostname match to node name? |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0016", variables => { node_name => $scan_cluster_node_name }}); |
|
next; |
|
} |
|
|
|
my $scan_cluster_node_pacemaker_id = $anvil->data->{cib}{parsed}{data}{node}{$scan_cluster_node_name}{node_state}{pacemaker_id}; |
|
my $scan_cluster_node_in_ccm = $anvil->data->{cib}{parsed}{data}{node}{$scan_cluster_node_name}{node_state}{in_ccm}; |
|
my $scan_cluster_node_crmd_member = $anvil->data->{cib}{parsed}{data}{node}{$scan_cluster_node_name}{node_state}{crmd}; |
|
my $scan_cluster_node_cluster_member = $anvil->data->{cib}{parsed}{data}{node}{$scan_cluster_node_name}{node_state}{'join'}; |
|
my $scan_cluster_node_maintenance_mode = $anvil->data->{cib}{parsed}{data}{node}{$scan_cluster_node_name}{node_state}{'maintenance-mode'}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
scan_cluster_node_pacemaker_id => $scan_cluster_node_pacemaker_id, |
|
scan_cluster_node_in_ccm => $scan_cluster_node_in_ccm, |
|
scan_cluster_node_crmd_member => $scan_cluster_node_crmd_member, |
|
scan_cluster_node_cluster_member => $scan_cluster_node_cluster_member, |
|
scan_cluster_node_maintenance_mode => $scan_cluster_node_maintenance_mode, |
|
}}); |
|
|
|
if (exists $anvil->data->{sql}{scan_cluster_node_host_uuid}{$scan_cluster_node_host_uuid}) |
|
{ |
|
# Look for changes. |
|
my $scan_cluster_node_uuid = $anvil->data->{sql}{scan_cluster_node_host_uuid}{$scan_cluster_node_host_uuid}; |
|
my $old_scan_cluster_node_name = $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_name}; |
|
my $old_scan_cluster_node_pacemaker_id = $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_pacemaker_id}; |
|
my $old_scan_cluster_node_in_ccm = $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_in_ccm}; |
|
my $old_scan_cluster_node_crmd_member = $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_crmd_member}; |
|
my $old_scan_cluster_node_cluster_member = $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_cluster_member}; |
|
my $old_scan_cluster_node_maintenance_mode = $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_maintenance_mode}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
scan_cluster_node_uuid => $scan_cluster_node_uuid, |
|
old_scan_cluster_node_name => $old_scan_cluster_node_name, |
|
old_scan_cluster_node_pacemaker_id => $old_scan_cluster_node_pacemaker_id, |
|
old_scan_cluster_node_in_ccm => $old_scan_cluster_node_in_ccm, |
|
old_scan_cluster_node_crmd_member => $old_scan_cluster_node_crmd_member, |
|
old_scan_cluster_node_cluster_member => $old_scan_cluster_node_cluster_member, |
|
old_scan_cluster_node_maintenance_mode => $old_scan_cluster_node_maintenance_mode, |
|
}}); |
|
|
|
my $update = 0; |
|
if ($scan_cluster_node_name ne $old_scan_cluster_node_name) |
|
{ |
|
$update = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { update => $update }}); |
|
|
|
my $variables = { |
|
new_node_name => $scan_cluster_node_name, |
|
old_node_name => $old_scan_cluster_node_name, |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0008", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0009", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
if ($scan_cluster_node_pacemaker_id ne $old_scan_cluster_node_pacemaker_id) |
|
{ |
|
$update = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { update => $update }}); |
|
|
|
my $variables = { |
|
node_name => $scan_cluster_node_name, |
|
new_pacemaker_id => $scan_cluster_node_pacemaker_id ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
old_pacemaker_id => $old_scan_cluster_node_pacemaker_id ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0004", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0004", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
if ($scan_cluster_node_in_ccm ne $old_scan_cluster_node_in_ccm) |
|
{ |
|
$update = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { update => $update }}); |
|
|
|
my $variables = { |
|
node_name => $scan_cluster_node_name, |
|
new_in_ccm => $scan_cluster_node_in_ccm ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
old_in_ccm => $old_scan_cluster_node_in_ccm ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0005", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0005", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
if ($scan_cluster_node_crmd_member ne $old_scan_cluster_node_crmd_member) |
|
{ |
|
$update = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { update => $update }}); |
|
|
|
my $variables = { |
|
node_name => $scan_cluster_node_name, |
|
new_crmd_member => $scan_cluster_node_crmd_member ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
old_crmd_member => $old_scan_cluster_node_crmd_member ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0006", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0006", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
if ($scan_cluster_node_cluster_member ne $old_scan_cluster_node_cluster_member) |
|
{ |
|
$update = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { update => $update }}); |
|
|
|
my $variables = { |
|
node_name => $scan_cluster_node_name, |
|
new_cluster_member => $scan_cluster_node_cluster_member ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
old_cluster_member => $old_scan_cluster_node_cluster_member ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0007", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0007", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
if ($scan_cluster_node_maintenance_mode ne $old_scan_cluster_node_maintenance_mode) |
|
{ |
|
$update = 1; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { update => $update }}); |
|
|
|
my $variables = { |
|
node_name => $scan_cluster_node_name, |
|
new_maintenance_mode => $scan_cluster_node_maintenance_mode ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
old_maintenance_mode => $old_scan_cluster_node_maintenance_mode ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0008", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0008", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { update => $update }}); |
|
if ($update) |
|
{ |
|
my $query = " |
|
UPDATE |
|
scan_cluster_nodes |
|
SET |
|
scan_cluster_node_name = ".$anvil->Database->quote($scan_cluster_node_name).", |
|
scan_cluster_node_pacemaker_id = ".$anvil->Database->quote($scan_cluster_node_pacemaker_id).", |
|
scan_cluster_node_in_ccm = ".$anvil->Database->quote($scan_cluster_node_in_ccm).", |
|
scan_cluster_node_crmd_member = ".$anvil->Database->quote($scan_cluster_node_crmd_member).", |
|
scan_cluster_node_cluster_member = ".$anvil->Database->quote($scan_cluster_node_cluster_member).", |
|
scan_cluster_node_maintenance_mode = ".$anvil->Database->quote($scan_cluster_node_maintenance_mode).", |
|
modified_date = ".$anvil->Database->quote($anvil->Database->refresh_timestamp)." |
|
WHERE |
|
scan_cluster_node_uuid = ".$anvil->Database->quote($scan_cluster_node_uuid)." |
|
;"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
$anvil->Database->write({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
} |
|
} |
|
else |
|
{ |
|
# Add the node. |
|
my $scan_cluster_node_uuid = $anvil->Get->uuid(); |
|
my $query = " |
|
INSERT INTO |
|
scan_cluster_nodes |
|
( |
|
scan_cluster_node_uuid, |
|
scan_cluster_node_scan_cluster_uuid, |
|
scan_cluster_node_host_uuid, |
|
scan_cluster_node_name, |
|
scan_cluster_node_pacemaker_id, |
|
scan_cluster_node_in_ccm, |
|
scan_cluster_node_crmd_member, |
|
scan_cluster_node_cluster_member, |
|
scan_cluster_node_maintenance_mode, |
|
modified_date |
|
) VALUES ( |
|
".$anvil->Database->quote($scan_cluster_node_uuid).", |
|
".$anvil->Database->quote($scan_cluster_uuid).", |
|
".$anvil->Database->quote($scan_cluster_node_host_uuid).", |
|
".$anvil->Database->quote($scan_cluster_node_name).", |
|
".$anvil->Database->quote($scan_cluster_node_pacemaker_id).", |
|
".$anvil->Database->quote($scan_cluster_node_in_ccm).", |
|
".$anvil->Database->quote($scan_cluster_node_crmd_member).", |
|
".$anvil->Database->quote($scan_cluster_node_cluster_member).", |
|
".$anvil->Database->quote($scan_cluster_node_maintenance_mode).", |
|
".$anvil->Database->quote($anvil->Database->refresh_timestamp)." |
|
);"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
$anvil->Database->write({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
|
|
my $host_name = $anvil->Get->host_name_from_uuid({host_uuid => $scan_cluster_node_host_uuid}); |
|
my $variables = { |
|
cluster_name => $cluster_name, |
|
node_name => $scan_cluster_node_name, |
|
host_uuid => $scan_cluster_node_host_uuid, |
|
host_name => $host_name, |
|
pacemaker_id => $scan_cluster_node_pacemaker_id, |
|
in_ccm => $scan_cluster_node_in_ccm ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", # Yes or No |
|
crmd_member => $scan_cluster_node_crmd_member ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
cluster_member => $scan_cluster_node_cluster_member ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
maintenance_mode => $scan_cluster_node_maintenance_mode ? "#!string!scan_cluster_unit_0001!#" : "#!string!scan_cluster_unit_0002!#", |
|
}; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0003", variables => $variables}); |
|
$anvil->Alert->register({debug => 2, alert_level => "notice", message => "scan_cluster_alert_0003", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
} |
|
|
|
### TODO: Check for / repair bad cluster config issues |
|
# If we're still alive, we're either node 1, or we're node 2 and node 1 is not ready. If we're not ready, |
|
if ($stonith_max_attempts ne "INFINITY") |
|
{ |
|
### TODO: Call pcs to update |
|
} |
|
|
|
return(0); |
|
} |
|
|
|
# Read in existing data from the database. |
|
sub read_last_scan |
|
{ |
|
my ($anvil) = @_; |
|
|
|
my $query = " |
|
SELECT |
|
scan_cluster_uuid, |
|
scan_cluster_anvil_uuid, |
|
scan_cluster_name, |
|
scan_cluster_cib |
|
FROM |
|
scan_cluster |
|
;"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
|
|
my $results = $anvil->Database->query({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
my $count = @{$results}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => { |
|
results => $results, |
|
count => $count, |
|
}}); |
|
foreach my $row (@{$results}) |
|
{ |
|
# NOTE: There's no known way to track a cluster name change, so we can't really avoid having |
|
# an entery per cluster name. |
|
my $scan_cluster_uuid = $row->[0]; |
|
my $scan_cluster_anvil_uuid = $row->[1]; |
|
my $scan_cluster_name = $row->[2]; |
|
my $scan_cluster_cib = $row->[3]; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
"s1:scan_cluster_uuid" => $scan_cluster_uuid, |
|
"s2:scan_cluster_anvil_uuid" => $scan_cluster_anvil_uuid, |
|
"s3:scan_cluster_name" => $scan_cluster_name, |
|
"s4:scan_cluster_cib" => $scan_cluster_cib, |
|
}}); |
|
|
|
# Store the old data now. |
|
$anvil->data->{sql}{scan_cluster}{scan_cluster_uuid}{$scan_cluster_uuid}{scan_cluster_name} = $scan_cluster_name; |
|
$anvil->data->{sql}{scan_cluster}{scan_cluster_uuid}{$scan_cluster_uuid}{scan_cluster_anvil_uuid} = $scan_cluster_anvil_uuid; |
|
$anvil->data->{sql}{scan_cluster}{scan_cluster_uuid}{$scan_cluster_uuid}{scan_cluster_cib} = $scan_cluster_cib; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
"sql::scan_cluster::scan_cluster_uuid::${scan_cluster_uuid}::scan_cluster_name" => $anvil->data->{sql}{scan_cluster}{scan_cluster_uuid}{$scan_cluster_uuid}{scan_cluster_name}, |
|
"sql::scan_cluster::scan_cluster_uuid::${scan_cluster_uuid}::scan_cluster_anvil_uuid" => $anvil->data->{sql}{scan_cluster}{scan_cluster_uuid}{$scan_cluster_uuid}{scan_cluster_anvil_uuid}, |
|
"sql::scan_cluster::scan_cluster_uuid::${scan_cluster_uuid}::scan_cluster_cib" => $anvil->data->{sql}{scan_cluster}{scan_cluster_uuid}{$scan_cluster_uuid}{scan_cluster_cib}, |
|
}}); |
|
|
|
# Make it easy to look up the cluster_uuid from the anvil_uuid. |
|
$anvil->data->{sql}{anvil_uuid}{$scan_cluster_anvil_uuid} = $scan_cluster_uuid; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
"sql::anvil_uuid::${scan_cluster_anvil_uuid}" => $anvil->data->{sql}{anvil_uuid}{$scan_cluster_anvil_uuid}, |
|
}}); |
|
} |
|
undef $count; |
|
undef $results; |
|
|
|
$query = " |
|
SELECT |
|
scan_cluster_node_uuid, |
|
scan_cluster_node_scan_cluster_uuid, |
|
scan_cluster_node_host_uuid, |
|
scan_cluster_node_name, |
|
scan_cluster_node_pacemaker_id, |
|
scan_cluster_node_in_ccm, |
|
scan_cluster_node_crmd_member, |
|
scan_cluster_node_cluster_member, |
|
scan_cluster_node_maintenance_mode |
|
FROM |
|
scan_cluster_nodes |
|
;"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
|
|
$results = $anvil->Database->query({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
$count = @{$results}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
results => $results, |
|
count => $count, |
|
}}); |
|
foreach my $row (@{$results}) |
|
{ |
|
# We've got an entry in the 'scan_cluster_nodes' table, so now we'll look for data in the node and |
|
# services tables. |
|
my $scan_cluster_node_uuid = $row->[0]; |
|
my $scan_cluster_node_scan_cluster_uuid = $row->[1]; |
|
my $scan_cluster_node_host_uuid = $row->[2]; |
|
my $scan_cluster_node_name = $row->[3]; |
|
my $scan_cluster_node_pacemaker_id = $row->[4]; |
|
my $scan_cluster_node_in_ccm = $row->[5]; |
|
my $scan_cluster_node_crmd_member = $row->[6]; |
|
my $scan_cluster_node_cluster_member = $row->[7]; |
|
my $scan_cluster_node_maintenance_mode = $row->[8]; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => { |
|
scan_cluster_node_uuid => $scan_cluster_node_uuid, |
|
scan_cluster_node_scan_cluster_uuid => $scan_cluster_node_scan_cluster_uuid, |
|
scan_cluster_node_host_uuid => $scan_cluster_node_host_uuid, |
|
scan_cluster_node_name => $scan_cluster_node_name, |
|
scan_cluster_node_pacemaker_id => $scan_cluster_node_pacemaker_id, |
|
scan_cluster_node_in_ccm => $scan_cluster_node_in_ccm, |
|
scan_cluster_node_crmd_member => $scan_cluster_node_crmd_member, |
|
scan_cluster_node_cluster_member => $scan_cluster_node_cluster_member, |
|
scan_cluster_node_maintenance_mode => $scan_cluster_node_maintenance_mode, |
|
}}); |
|
|
|
# Store the old data now. |
|
$anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid} = { |
|
scan_cluster_node_scan_cluster_uuid => $scan_cluster_node_scan_cluster_uuid, |
|
scan_cluster_node_host_uuid => $scan_cluster_node_host_uuid, |
|
scan_cluster_node_name => $scan_cluster_node_name, |
|
scan_cluster_node_pacemaker_id => $scan_cluster_node_pacemaker_id, |
|
scan_cluster_node_in_ccm => $scan_cluster_node_in_ccm, |
|
scan_cluster_node_crmd_member => $scan_cluster_node_crmd_member, |
|
scan_cluster_node_cluster_member => $scan_cluster_node_cluster_member, |
|
scan_cluster_node_maintenance_mode => $scan_cluster_node_maintenance_mode, |
|
}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
"sql::scan_cluster_node::scan_cluster_node_uuid::${scan_cluster_node_uuid}::scan_cluster_node_scan_cluster_uuid" => $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_scan_cluster_uuid}, |
|
"sql::scan_cluster_node::scan_cluster_node_uuid::${scan_cluster_node_uuid}::scan_cluster_node_host_uuid" => $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_host_uuid}, |
|
"sql::scan_cluster_node::scan_cluster_node_uuid::${scan_cluster_node_uuid}::scan_cluster_node_name" => $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_name}, |
|
"sql::scan_cluster_node::scan_cluster_node_uuid::${scan_cluster_node_uuid}::scan_cluster_node_pacemaker_id" => $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_pacemaker_id}, |
|
"sql::scan_cluster_node::scan_cluster_node_uuid::${scan_cluster_node_uuid}::scan_cluster_node_in_ccm" => $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_in_ccm}, |
|
"sql::scan_cluster_node::scan_cluster_node_uuid::${scan_cluster_node_uuid}::scan_cluster_node_crmd_member" => $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_crmd_member}, |
|
"sql::scan_cluster_node::scan_cluster_node_uuid::${scan_cluster_node_uuid}::scan_cluster_node_cluster_member" => $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_cluster_member}, |
|
"sql::scan_cluster_node::scan_cluster_node_uuid::${scan_cluster_node_uuid}::scan_cluster_node_maintenance_mode" => $anvil->data->{sql}{scan_cluster_node}{scan_cluster_node_uuid}{$scan_cluster_node_uuid}{scan_cluster_node_maintenance_mode}, |
|
}}); |
|
|
|
$anvil->data->{sql}{scan_cluster_node_host_uuid}{$scan_cluster_node_host_uuid} = $scan_cluster_node_uuid; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
"sql::scan_cluster_node_host_uuid::${scan_cluster_node_host_uuid}" => $anvil->data->{sql}{scan_cluster_node_host_uuid}{$scan_cluster_node_host_uuid}, |
|
}}); |
|
} |
|
|
|
return(0); |
|
} |
|
|
|
# This reads in all the data we can find on the local system |
|
sub collect_data |
|
{ |
|
my ($anvil) = @_; |
|
|
|
# Pick out core cluster details. |
|
my $problem = $anvil->Cluster->parse_cib({debug => 2}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { problem => $problem }}); |
|
|
|
# If there was a problem, we're not in the cluster. |
|
if ($problem) |
|
{ |
|
my $changed = $anvil->Alert->check_alert_sent({ |
|
record_locator => "scan_cluster::in_cluster", |
|
set_by => $THIS_FILE, |
|
}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { changed => $changed }}); |
|
if ($changed) |
|
{ |
|
# Register an alert. |
|
my $variables = { host_name => $anvil->Get->host_name() }; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0010", variables => $variables}); |
|
$anvil->Alert->register({alert_level => "warning", message => "scan_cluster_alert_0010", variables => $variables, set_by => $THIS_FILE}); |
|
|
|
# See if I need to mark us as out of the cluster. Normally, our peer would do this, |
|
# but if we went down at the same time as our peer, both of us might not update the |
|
# membership values. |
|
my $query = " |
|
SELECT |
|
scan_cluster_node_uuid, |
|
scan_cluster_node_in_ccm, |
|
scan_cluster_node_crmd_member, |
|
scan_cluster_node_cluster_member |
|
FROM |
|
scan_cluster_nodes |
|
WHERE |
|
scan_cluster_node_host_uuid = ".$anvil->Database->quote($anvil->Get->host_uuid)." |
|
;"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
|
|
my $results = $anvil->Database->query({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
my $count = @{$results}; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
results => $results, |
|
count => $count, |
|
}}); |
|
foreach my $row (@{$results}) |
|
{ |
|
# We've got an entry in the 'scan_cluster_nodes' table, so now we'll look for data in the node and |
|
# services tables. |
|
my $scan_cluster_node_uuid = $row->[0]; |
|
my $scan_cluster_node_in_ccm = $row->[1]; |
|
my $scan_cluster_node_crmd_member = $row->[2]; |
|
my $scan_cluster_node_cluster_member = $row->[3]; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { |
|
scan_cluster_node_uuid => $scan_cluster_node_uuid, |
|
scan_cluster_node_in_ccm => $scan_cluster_node_in_ccm, |
|
scan_cluster_node_crmd_member => $scan_cluster_node_crmd_member, |
|
scan_cluster_node_cluster_member => $scan_cluster_node_cluster_member, |
|
}}); |
|
|
|
if (($scan_cluster_node_in_ccm) or ($scan_cluster_node_crmd_member) or ($scan_cluster_node_cluster_member)) |
|
{ |
|
# Update |
|
my $query = " |
|
UPDATE |
|
scan_cluster_nodes |
|
SET |
|
scan_cluster_node_in_ccm = '0', |
|
scan_cluster_node_crmd_member = '0', |
|
scan_cluster_node_cluster_member = '0', |
|
modified_date = ".$anvil->Database->quote($anvil->Database->refresh_timestamp)." |
|
WHERE |
|
scan_cluster_node_uuid = ".$anvil->Database->quote($scan_cluster_node_uuid)." |
|
;"; |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }}); |
|
$anvil->Database->write({query => $query, source => $THIS_FILE, line => __LINE__}); |
|
} |
|
} |
|
} |
|
|
|
# Exit now. |
|
$anvil->nice_exit({exit_code => 2}); |
|
} |
|
else |
|
{ |
|
# See if we came back into the cluster |
|
my $changed = $anvil->Alert->check_alert_sent({ |
|
clear => 1, |
|
record_locator => "scan_cluster::in_cluster", |
|
set_by => $THIS_FILE, |
|
}); |
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { changed => $changed }}); |
|
if ($changed) |
|
{ |
|
# Register an alert. |
|
my $variables = { host_name => $anvil->Get->host_name() }; |
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_cluster_alert_0011", variables => $variables}); |
|
$anvil->Alert->register({alert_level => "warning", clear_alert => 1, message => "scan_cluster_alert_0011", variables => $variables, set_by => $THIS_FILE}); |
|
} |
|
} |
|
|
|
return(0); |
|
}
|
|
|