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.
716 lines
31 KiB
716 lines
31 KiB
3 years ago
|
#!/usr/bin/perl
|
||
|
|
#
|
||
|
|
# This manages if a server is backed up to a DR host or not. When enabled, it can start or stop replication.
|
||
|
|
#
|
||
|
|
# NOTE: Unlike most jobs, this one will directly work on the peer node and the DR host using SSH connections.
|
||
|
|
# This behaviour is likely to change later as it's not ideal.
|
||
|
|
#
|
||
|
|
# Exit codes;
|
||
|
|
# 0 = Normal exit.
|
||
|
|
# 1 = Any problem that causes an early exit.
|
||
|
|
#
|
||
|
|
|
||
|
|
use strict;
|
||
|
|
use warnings;
|
||
|
|
use Anvil::Tools;
|
||
|
|
require POSIX;
|
||
|
|
use Term::Cap;
|
||
|
|
|
||
|
|
my $THIS_FILE = ($0 =~ /^.*\/(.*)$/)[0];
|
||
|
|
my $running_directory = ($0 =~ /^(.*?)\/$THIS_FILE$/)[0];
|
||
|
|
if (($running_directory =~ /^\./) && ($ENV{PWD}))
|
||
|
|
{
|
||
|
|
$running_directory =~ s/^\./$ENV{PWD}/;
|
||
|
|
}
|
||
|
|
|
||
|
|
# Turn off buffering so that the pinwheel will display while waiting for the SSH call(s) to complete.
|
||
|
|
$| = 1;
|
||
|
|
|
||
|
|
my $anvil = Anvil::Tools->new();
|
||
|
|
|
||
|
|
#
|
||
|
|
$anvil->data->{switches}{'connect'} = ""; # connect an existing DR resource
|
||
|
|
$anvil->data->{switches}{disconnect} = ""; # disconnect
|
||
|
|
$anvil->data->{switches}{'job-uuid'} = ""; # Used later
|
||
|
|
$anvil->data->{switches}{protect} = ""; # Set
|
||
|
|
$anvil->data->{switches}{protocol} = ""; # "sync", "async" or "long-throw"
|
||
|
|
$anvil->data->{switches}{remove} = ""; # Set
|
||
|
|
$anvil->data->{switches}{server} = ""; # Name or UUID
|
||
|
|
$anvil->data->{switches}{update} = ""; # connects, if needed, and disconnects once UpToDate
|
||
|
|
$anvil->data->{switches}{Yes} = ""; # Set to avoid confirmation, not case sensitive
|
||
|
|
$anvil->Get->switches;
|
||
|
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 2, secure => 0, key => "log_0115", variables => { program => $THIS_FILE }});
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
'switches::connect' => $anvil->data->{switches}{'connect'},
|
||
|
|
'switches::disconnect' => $anvil->data->{switches}{disconnect},
|
||
|
|
'switches::job-uuid' => $anvil->data->{switches}{'job-uuid'},
|
||
|
|
'switches::protect' => $anvil->data->{switches}{protect},
|
||
|
|
'switches::protocol' => $anvil->data->{switches}{protocol},
|
||
|
|
'switches::remove' => $anvil->data->{switches}{remove},
|
||
|
|
'switches::server' => $anvil->data->{switches}{server},
|
||
|
|
'switches::update' => $anvil->data->{switches}{update},
|
||
|
|
'switches::Yes' => $anvil->data->{switches}{Yes},
|
||
|
|
}});
|
||
|
|
|
||
|
|
$anvil->Database->connect();
|
||
|
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 3, secure => 0, key => "log_0132"});
|
||
|
|
if (not $anvil->data->{sys}{database}{connections})
|
||
|
|
{
|
||
|
|
# No databases, update the job, sleep for a bit and then exit. The daemon will pick it up and try
|
||
|
|
# again after we exit.
|
||
|
|
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, 'print' => 1, level => 0, priority => "err", key => "error_0306"});
|
||
|
|
sleep 10;
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
# If we've got a job UUID, load the job details.
|
||
|
|
if ($anvil->data->{switches}{'job-uuid'})
|
||
|
|
{
|
||
|
|
load_job($anvil);
|
||
|
|
}
|
||
|
|
|
||
|
|
sanity_check($anvil);
|
||
|
|
|
||
|
|
do_task($anvil);
|
||
|
|
|
||
|
|
|
||
|
|
$anvil->nice_exit({exit_code => 0});
|
||
|
|
|
||
|
|
|
||
|
|
#############################################################################################################
|
||
|
|
# Functions #
|
||
|
|
#############################################################################################################
|
||
|
|
|
||
|
|
sub do_task
|
||
|
|
{
|
||
|
|
my ($anvil) = @_;
|
||
|
|
|
||
|
|
# What task am I doing?
|
||
|
|
if ($anvil->data->{switches}{protect})
|
||
|
|
{
|
||
|
|
|
||
|
|
}
|
||
|
|
|
||
|
|
return(0);
|
||
|
|
}
|
||
|
|
|
||
|
|
sub sanity_check
|
||
|
|
{
|
||
|
|
my ($anvil) = @_;
|
||
|
|
|
||
|
|
# Are we a node or DR?
|
||
|
|
my $host_type = $anvil->Get->host_type();
|
||
|
|
my $anvil_uuid = $anvil->Cluster->get_anvil_uuid();
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
host_type => $host_type,
|
||
|
|
anvil_uuid => $anvil_uuid,
|
||
|
|
}});
|
||
|
|
|
||
|
|
if (($host_type ne "node") or (not $anvil_uuid))
|
||
|
|
{
|
||
|
|
print "This must be run on a node active in the cluster hosting the server being managed. Exiting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
# Get the Anvil! details.
|
||
|
|
$anvil->Database->get_hosts();
|
||
|
|
$anvil->Database->get_anvils();
|
||
|
|
$anvil->Database->get_storage_group_data({debug => 2});
|
||
|
|
|
||
|
|
# Does this Anvil! have a DR node?
|
||
|
|
if (not $anvil->data->{anvils}{anvil_uuid}{$anvil_uuid}{anvil_dr1_host_uuid})
|
||
|
|
{
|
||
|
|
print "This Anvil! does not seem to have a DR host. Exiting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
# Can we access DR?
|
||
|
|
my $password = $anvil->data->{anvils}{anvil_uuid}{$anvil_uuid}{anvil_password};
|
||
|
|
my $dr_host_uuid = $anvil->data->{anvils}{anvil_uuid}{$anvil_uuid}{anvil_dr1_host_uuid};
|
||
|
|
my $dr_host_name = $anvil->data->{hosts}{host_uuid}{$dr_host_uuid}{host_name};
|
||
|
|
my $dr_ip = $anvil->System->find_matching_ip({
|
||
|
|
debug => 2,
|
||
|
|
host => $dr_host_name,
|
||
|
|
});
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
password => $anvil->Log->is_secure($password),
|
||
|
|
dr_host_uuid => $dr_host_uuid,
|
||
|
|
dr_host_name => $dr_host_name,
|
||
|
|
dr_ip => $dr_ip,
|
||
|
|
}});
|
||
|
|
if ((not $dr_ip) or ($dr_ip eq "!!error!!"))
|
||
|
|
{
|
||
|
|
print "Failed to find an IP we can access the DR host: [".$dr_host_name."]. Has it been configured? Is it running? Exiting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
# Test access.
|
||
|
|
my $access = $anvil->Remote->test_access({
|
||
|
|
target => $dr_ip,
|
||
|
|
password => $password,
|
||
|
|
});
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { access => $access }});
|
||
|
|
if (not $access)
|
||
|
|
{
|
||
|
|
print "Failed to access the DR host: [".$dr_host_name."] using the IP: [".$dr_ip."]. Is it running? Exiting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
# Can we parse the CIB?
|
||
|
|
my ($problem) = $anvil->Cluster->parse_cib();
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { problem => $problem }});
|
||
|
|
if ($problem)
|
||
|
|
{
|
||
|
|
print "Failed to parse the CIB. Is this node in the cluster? Exiting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
# Both nodes need to be in the cluster, are they?
|
||
|
|
if (not $anvil->data->{cib}{parsed}{'local'}{ready})
|
||
|
|
{
|
||
|
|
print "We're not a full member of the cluster yet. Please try again once we're fully in. Exiting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
### TODO: We can queue a job to update the peer later, there's no real need, in the long run, for the
|
||
|
|
### peer to be online.
|
||
|
|
# If we're protecting or removing a server from DR, the peer needs to be up.
|
||
|
|
if ((($anvil->data->{switches}{protect}) or
|
||
|
|
($anvil->data->{switches}{remove}) or
|
||
|
|
($anvil->data->{switches}{protocol})) &&
|
||
|
|
(not $anvil->data->{cib}{parsed}{peer}{ready}))
|
||
|
|
{
|
||
|
|
if ($anvil->data->{switches}{protect})
|
||
|
|
{
|
||
|
|
print "We can't setup a server to be protected unless both nodes are up, and the peer isn't at this time. Exiting.\n";
|
||
|
|
}
|
||
|
|
else
|
||
|
|
{
|
||
|
|
print "We can't remove a server from DR unless both nodes are up, and the peer isn't at this time. Exiting.\n";
|
||
|
|
}
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
# Verify we found the server.
|
||
|
|
$anvil->data->{server}{'server-name'} = "";
|
||
|
|
$anvil->data->{server}{'server-uuid'} = "";
|
||
|
|
$anvil->data->{server}{'anvil-uuid'} = $anvil_uuid;
|
||
|
|
if (not $anvil->data->{switches}{server})
|
||
|
|
{
|
||
|
|
print "Please specify the server to manager using '--server <name or uuid>'. Exiting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
else
|
||
|
|
{
|
||
|
|
my $server = $anvil->data->{switches}{server};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { server => $server }});
|
||
|
|
|
||
|
|
$anvil->Database->get_servers();
|
||
|
|
if (exists $anvil->data->{servers}{server_uuid}{$server})
|
||
|
|
{
|
||
|
|
$anvil->data->{server}{'server-uuid'} = $server;
|
||
|
|
$anvil->data->{server}{'server-name'} = $anvil->data->{servers}{server_uuid}{$server}{server_name};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
'server::server-uuid' => $anvil->data->{server}{'server-uuid'},
|
||
|
|
'server::server-name' => $anvil->data->{server}{'server-name'},
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
if (exists $anvil->data->{servers}{anvil_uuid}{$anvil_uuid}{server_name}{$server})
|
||
|
|
{
|
||
|
|
$anvil->data->{server}{'server-name'} = $server;
|
||
|
|
$anvil->data->{server}{'server-uuid'} = $anvil->data->{servers}{anvil_uuid}{$anvil_uuid}{server_name}{$server}{server_uuid};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
'server::server-uuid' => $anvil->data->{server}{'server-uuid'},
|
||
|
|
'server::server-name' => $anvil->data->{server}{'server-name'},
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
# Get and parse the server's definition to find the DRBD devices.
|
||
|
|
if ((not $anvil->data->{server}{'server-uuid'}) or (not $anvil->data->{server}{'server-name'}))
|
||
|
|
{
|
||
|
|
print "Failed to find the server: [".$anvil->data->{switches}{server}."] by name or UUID? Exiting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
if (not $anvil->data->{switches}{protocol})
|
||
|
|
{
|
||
|
|
$anvil->data->{switches}{protocol} = "async";
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
'switches::protocol' => $anvil->data->{switches}{protocol},
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
elsif (($anvil->data->{switches}{protocol} ne "sync") &&
|
||
|
|
($anvil->data->{switches}{protocol} ne "async") &&
|
||
|
|
($anvil->data->{switches}{protocol} ne "long-throw"))
|
||
|
|
{
|
||
|
|
print "The protocol: [".$anvil->data->{switches}{protocol}."] is invalid. Please use '--help' for more information.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
# Are we being asked to actuall do something?
|
||
|
|
if (((not $anvil->data->{switches}{'connect'}) &&
|
||
|
|
(not $anvil->data->{switches}{disconnect}) &&
|
||
|
|
(not $anvil->data->{switches}{protect}) &&
|
||
|
|
(not $anvil->data->{switches}{remove}) &&
|
||
|
|
(not $anvil->data->{switches}{update})) or
|
||
|
|
($anvil->data->{switches}{help}) or
|
||
|
|
($anvil->data->{switches}{h}))
|
||
|
|
{
|
||
|
|
print "
|
||
|
|
What do you want to do?
|
||
|
|
|
||
|
|
Options (all require --server <name or UUID>);
|
||
|
|
|
||
|
|
--connect
|
||
|
|
|
||
|
|
Connect a server already on DR to it's DR copy, update the data there if needed and begin streaming
|
||
|
|
replication.
|
||
|
|
|
||
|
|
--disconnect
|
||
|
|
|
||
|
|
Disconnect a server from the DR image. This will end streaming replication.
|
||
|
|
|
||
|
|
--protect
|
||
|
|
|
||
|
|
The sets up the server to be imaged on DR, if it isn't already protected.
|
||
|
|
|
||
|
|
Notes: If the server is not running, the DRBD resource volume(s) will be brought up. Both nodes need
|
||
|
|
to be online and in the cluster.
|
||
|
|
|
||
|
|
--protocol <sync,async,long-throw>, default 'async'
|
||
|
|
|
||
|
|
This allows the protocol used to replicate data to the DR host to be configured. By default, 'async'
|
||
|
|
is used.
|
||
|
|
|
||
|
|
Modes:
|
||
|
|
|
||
|
|
async (default)
|
||
|
|
|
||
|
|
This tells the storage layer to consider the write to be completed once the data is on the
|
||
|
|
active node's network transmit buffer. In this way, the DR host is allowed to fall behind a
|
||
|
|
small amount, but the active nodes will not slow down because of higher network transit times
|
||
|
|
to the DR location.
|
||
|
|
|
||
|
|
NOTE: The transmit (TX) buffer size can be checked / updated with 'ethtool -g <link_device>'.
|
||
|
|
If the transmit buffer fills, storage will hold until the buffer flushes, causing
|
||
|
|
periodic storage IO waits. You can increase the buffer size to a certain degree with
|
||
|
|
'ethtool -G <link_device> tx <size>' (set on all storage network link devices on both
|
||
|
|
nodes. For more information, see:
|
||
|
|
|
||
|
|
https://www.linuxjournal.com/content/queueing-linux-network-stack
|
||
|
|
|
||
|
|
or
|
||
|
|
|
||
|
|
https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/configuring_and_managing_networking/monitoring-and-tuning-the-rx-ring-buffer_configuring-and-managing-networking
|
||
|
|
|
||
|
|
If you set the maximum transmit buffer size and still run into IO waits, consider
|
||
|
|
'long-throw'.
|
||
|
|
|
||
|
|
sync
|
||
|
|
|
||
|
|
This tells the storage layer to consider the write complete when the data has reached the DR
|
||
|
|
host's storage (when the data is committed to disk on DR). This means that the DR host will
|
||
|
|
never fall behind. However, if the DR's network latency is higher or the bandwidth to the DR
|
||
|
|
is lower than that of the latency/bandwidth between the nodes, then total storage performance
|
||
|
|
will be reduced to DR network speeds while DR is connected.
|
||
|
|
|
||
|
|
This should be tested before implemented in production.
|
||
|
|
|
||
|
|
long-throw
|
||
|
|
|
||
|
|
This is an option that requires an additional license fee to use.
|
||
|
|
|
||
|
|
This option (based on LINBIT's DRBD Proxy) and is designed for DR hosts that are connected
|
||
|
|
over a wide-area network (or other cases where the connection to the DR is high-latency, low
|
||
|
|
bandwidth or intermittently interrupted). It uses RAM on the host to act, effectively, as a
|
||
|
|
very large transmit buffer. This requires allocating host RAM to the task, and so could
|
||
|
|
reduces the available RAM assignable to assign to servers.
|
||
|
|
|
||
|
|
In this mode, the DR host is allowed to fall further behind production, but it significantly
|
||
|
|
reduces (hopefully eliminates) how often node replication waits because of a full transmit
|
||
|
|
buffer.
|
||
|
|
|
||
|
|
The default size is 16 MiB, with a maximum size of 16 GiB. When the size is set to over
|
||
|
|
1 GiB, the size allocated to this buffer is accounted for when calculating available RAM that
|
||
|
|
can be assigned to hosted servers.
|
||
|
|
|
||
|
|
--remove
|
||
|
|
|
||
|
|
This removes the DR image from the DR host for the server, freeing up space on DR but removing the
|
||
|
|
protection afforded by DR.
|
||
|
|
|
||
|
|
--update
|
||
|
|
|
||
|
|
This tells the DR to be connected and sync, Once the volume(s) on DR are 'UpToDate', the connection
|
||
|
|
is closed. This provides a point in time update of the server's image on DR.
|
||
|
|
|
||
|
|
--Yes
|
||
|
|
|
||
|
|
Note the capital 'Y'. This can be set to proceed without confirmation. Use carefully with '--protect'
|
||
|
|
and '--remove'! If the '--job-uuid' is set, this is assumed and no prompt will be presented.
|
||
|
|
|
||
|
|
Exiting.
|
||
|
|
";
|
||
|
|
if (($anvil->data->{switches}{help}) or ($anvil->data->{switches}{h}))
|
||
|
|
{
|
||
|
|
$anvil->nice_exit({exit_code => 0});
|
||
|
|
}
|
||
|
|
else
|
||
|
|
{
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
# If we're protecting, make sure there's enough space on the DR host.
|
||
|
|
if ($anvil->data->{switches}{protect})
|
||
|
|
{
|
||
|
|
prepare_for_protect($anvil);
|
||
|
|
}
|
||
|
|
|
||
|
|
return(0);
|
||
|
|
}
|
||
|
|
|
||
|
|
sub prepare_for_protect
|
||
|
|
{
|
||
|
|
my ($anvil) = @_;
|
||
|
|
|
||
|
|
# Parse out the DRBD resource's backing the server and get their LV sizes.
|
||
|
|
$anvil->Database->get_server_definitions();
|
||
|
|
my $anvil_uuid = $anvil->Cluster->get_anvil_uuid();
|
||
|
|
my $dr_host_uuid = $anvil->data->{anvils}{anvil_uuid}{$anvil_uuid}{anvil_dr1_host_uuid};
|
||
|
|
my $dr_host_name = $anvil->data->{hosts}{host_uuid}{$dr_host_uuid}{host_name};
|
||
|
|
my $server_name = $anvil->data->{server}{'server-name'};
|
||
|
|
my $server_uuid = $anvil->data->{server}{'server-uuid'};
|
||
|
|
my $short_host_name = $anvil->Get->short_host_name();
|
||
|
|
my $server_definition_xml = $anvil->data->{server_definitions}{server_definition_server_uuid}{$server_uuid}{server_definition_xml};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
anvil_uuid => $anvil_uuid,
|
||
|
|
dr_host_uuid => $dr_host_uuid,
|
||
|
|
dr_host_name => $dr_host_name,
|
||
|
|
server_name => $server_name,
|
||
|
|
server_uuid => $server_uuid,
|
||
|
|
server_definition_xml => $server_definition_xml,
|
||
|
|
short_host_name => $short_host_name,
|
||
|
|
}});
|
||
|
|
|
||
|
|
$anvil->Server->parse_definition({
|
||
|
|
debug => 2,
|
||
|
|
host => $short_host_name,
|
||
|
|
server => $anvil->data->{server}{'server-name'},
|
||
|
|
source => "from_db",
|
||
|
|
definition => $server_definition_xml,
|
||
|
|
});
|
||
|
|
|
||
|
|
$anvil->DRBD->gather_data();
|
||
|
|
|
||
|
|
my $server_ram = $anvil->data->{server}{$short_host_name}{$server_name}{'from_db'}{memory};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
server_ram => $anvil->Convert->add_commas({number => $server_ram})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $server_ram}).")",
|
||
|
|
}});
|
||
|
|
foreach my $resource (sort {$a cmp $b} keys %{$anvil->data->{server}{$short_host_name}{$server_name}{drbd}{resource}})
|
||
|
|
{
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { resource => $resource }});
|
||
|
|
|
||
|
|
foreach my $this_host_name (sort {$a cmp $b} keys %{$anvil->data->{new}{resource}{$resource}{host}})
|
||
|
|
{
|
||
|
|
my $this_host_uuid = $anvil->Get->host_uuid_from_name({host_name => $this_host_name});
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
this_host_name => $this_host_name,
|
||
|
|
this_host_uuid => $this_host_uuid,
|
||
|
|
}});
|
||
|
|
|
||
|
|
foreach my $volume (sort {$a cmp $b} keys %{$anvil->data->{new}{resource}{$resource}{host}{$this_host_name}{volume}})
|
||
|
|
{
|
||
|
|
# Always get the LV sizes, as that factors metadata. DRBD size is
|
||
|
|
# minus metadata, and 0 when down.
|
||
|
|
my $device_path = $anvil->data->{new}{resource}{$resource}{host}{$this_host_name}{volume}{$volume}{device_path};
|
||
|
|
my $backing_disk = $anvil->data->{new}{resource}{$resource}{host}{$this_host_name}{volume}{$volume}{backing_disk};
|
||
|
|
my $device_minor = $anvil->data->{new}{resource}{$resource}{host}{$this_host_name}{volume}{$volume}{device_minor};
|
||
|
|
my $tcp_port = $anvil->data->{new}{resource}{$resource}{peer}{$this_host_name}{tcp_port};
|
||
|
|
my $this_size = $anvil->Storage->get_size_of_block_device({host_uuid => $this_host_uuid, path => $backing_disk});
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"s1:volume" => $volume,
|
||
|
|
"s2:device_path" => $device_path,
|
||
|
|
"s3:backing_disk" => $backing_disk,
|
||
|
|
"s4:device_minor" => $device_minor,
|
||
|
|
"s5:this_size" => $anvil->Convert->add_commas({number => $this_size})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $this_size}).")",
|
||
|
|
}});
|
||
|
|
|
||
|
|
if ((not exists $anvil->data->{server}{drbd}{$resource}{$volume}{size}) or (not $anvil->data->{server}{drbd}{$resource}{$volume}{size}))
|
||
|
|
{
|
||
|
|
$anvil->data->{server}{drbd}{$resource}{$volume}{size} = $this_size;
|
||
|
|
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"server::drbd::${resource}::${volume}::size" => $anvil->data->{server}{drbd}{$resource}{$volume}{size},
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
|
||
|
|
if (not exists $anvil->data->{server}{drbd}{$resource}{$volume}{storage_group_uuid})
|
||
|
|
{
|
||
|
|
$anvil->data->{server}{drbd}{$resource}{$volume}{storage_group_uuid} = "";
|
||
|
|
}
|
||
|
|
|
||
|
|
### NOTE: This check make sense only under the assumption that the DRBD minor
|
||
|
|
### is common across both nodes. This should be the case, but doesn't
|
||
|
|
### strictly have to be so.
|
||
|
|
if ((not exists $anvil->data->{server}{drbd}{$resource}{$volume}{minor_number}) or
|
||
|
|
(not defined $anvil->data->{server}{drbd}{$resource}{$volume}{minor_number}) or
|
||
|
|
($anvil->data->{server}{drbd}{$resource}{$volume}{minor_number} eq ""))
|
||
|
|
{
|
||
|
|
$anvil->data->{server}{drbd}{$resource}{$volume}{minor_number} = $device_minor;
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"server::drbd::${resource}::${volume}::minor_number" => $anvil->data->{server}{drbd}{$resource}{$volume}{minor_number},
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
|
||
|
|
if ((not exists $anvil->data->{server}{drbd}{$resource}{$volume}{tcp_port}) or
|
||
|
|
(not defined $anvil->data->{server}{drbd}{$resource}{$volume}{tcp_port}) or
|
||
|
|
($anvil->data->{server}{drbd}{$resource}{$volume}{tcp_port} eq ""))
|
||
|
|
{
|
||
|
|
$anvil->data->{server}{drbd}{$resource}{$volume}{tcp_port} = $tcp_port;
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"server::drbd::${resource}::${volume}::tcp_port" => $anvil->data->{server}{drbd}{$resource}{$volume}{tcp_port},
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
|
||
|
|
# What storage group does this belong to?
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"server::drbd::${resource}::${volume}::storage_group_uuid" => $anvil->data->{server}{drbd}{$resource}{$volume}{storage_group_uuid},
|
||
|
|
}});
|
||
|
|
if (not $anvil->data->{server}{drbd}{$resource}{$volume}{storage_group_uuid})
|
||
|
|
{
|
||
|
|
my $storage_key = $resource."/".$volume;
|
||
|
|
my $storage_group_uuid = $anvil->Storage->get_storage_group_from_path({
|
||
|
|
debug => 2,
|
||
|
|
anvil_uuid => $anvil_uuid,
|
||
|
|
path => $backing_disk,
|
||
|
|
});
|
||
|
|
my $storage_group_name = $anvil->data->{storage_groups}{anvil_uuid}{$anvil_uuid}{storage_group_uuid}{$storage_group_uuid}{group_name};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
storage_key => $storage_key,
|
||
|
|
storage_group_uuid => $storage_group_uuid,
|
||
|
|
storage_group_name => $storage_group_name,
|
||
|
|
}});
|
||
|
|
|
||
|
|
# We'll need to sum up the volumes on each storage group, as
|
||
|
|
# it's possible the volumes are on different SGs.
|
||
|
|
$anvil->data->{server}{drbd}{$resource}{$volume}{storage_group_uuid} = $storage_group_uuid;
|
||
|
|
$anvil->data->{server}{storage_groups}{$storage_group_name}{used_by}{$storage_key} = 1;
|
||
|
|
$anvil->data->{server}{storage_groups}{$storage_group_name}{storage_group_uuid} = $storage_group_uuid;
|
||
|
|
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"server::drbd::${resource}::${volume}::storage_group_uuid" => $anvil->data->{server}{drbd}{$resource}{$volume}{storage_group_uuid},
|
||
|
|
"server::storage_groups::${storage_group_name}::used_by::${storage_key}" => $anvil->data->{server}{storage_groups}{$storage_group_name}{used_by}{$storage_key},
|
||
|
|
"server::storage_groups::${storage_group_name}::storage_group_uuid" => $anvil->data->{server}{storage_groups}{$storage_group_name}{storage_group_uuid},
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
|
||
|
|
if ($this_size > $anvil->data->{server}{drbd}{$resource}{$volume}{size})
|
||
|
|
{
|
||
|
|
$anvil->data->{server}{drbd}{$resource}{$volume}{size} = $this_size;
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"server::drbd::${resource}::${volume}::size" => $anvil->Convert->add_commas({number => $anvil->data->{server}{drbd}{$resource}{$volume}{size}})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $anvil->data->{server}{drbd}{$resource}{$volume}{size}}).")",
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
# Make sure there is enough space on DR for the volumes under this VM.
|
||
|
|
my $problem = 0;
|
||
|
|
foreach my $storage_group_name (sort {$a cmp $b} keys %{$anvil->data->{server}{storage_groups}})
|
||
|
|
{
|
||
|
|
my $storage_group_uuid = $anvil->data->{server}{storage_groups}{$storage_group_name}{storage_group_uuid};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
storage_group_name => $storage_group_name,
|
||
|
|
storage_group_uuid => $storage_group_uuid,
|
||
|
|
}});
|
||
|
|
|
||
|
|
# First, is this SG on DR?
|
||
|
|
if (not exists $anvil->data->{storage_groups}{anvil_uuid}{$anvil_uuid}{storage_group_uuid}{$storage_group_uuid}{host_uuid}{$dr_host_uuid})
|
||
|
|
{
|
||
|
|
print "The DR host: [".$dr_host_name."] doesn't appear to be storage group: [".$storage_group_name."]. Unable to proceed.\n";
|
||
|
|
$problem = 1;
|
||
|
|
}
|
||
|
|
|
||
|
|
my $space_needed = 0;
|
||
|
|
foreach my $resource_key (sort {$a cmp $b} keys %{$anvil->data->{server}{storage_groups}{$storage_group_name}{used_by}})
|
||
|
|
{
|
||
|
|
my ($resource, $volume) = ($resource_key =~ /^(.*)\/(\d+)$/);
|
||
|
|
my $volume_size = $anvil->data->{server}{drbd}{$resource}{$volume}{size};
|
||
|
|
$space_needed += $volume_size,
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
resource_key => $resource_key,
|
||
|
|
resource => $resource,
|
||
|
|
volume => $volume,
|
||
|
|
volume_size => $anvil->Convert->add_commas({number => $volume_size})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $volume_size}).")",
|
||
|
|
space_needed => $anvil->Convert->add_commas({number => $space_needed})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $space_needed}).")",
|
||
|
|
}});
|
||
|
|
|
||
|
|
}
|
||
|
|
|
||
|
|
# Is there enough space on DR?
|
||
|
|
my $space_on_dr = $anvil->data->{storage_groups}{anvil_uuid}{$anvil_uuid}{storage_group_uuid}{$storage_group_uuid}{host_uuid}{$dr_host_uuid}{vg_free};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
space_on_dr => $anvil->Convert->add_commas({number => $space_on_dr})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $space_on_dr}).")",
|
||
|
|
space_needed => $anvil->Convert->add_commas({number => $space_needed})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $space_needed}).")",
|
||
|
|
}});
|
||
|
|
if ($space_needed > $space_on_dr)
|
||
|
|
{
|
||
|
|
print "We need: [".$anvil->Convert->bytes_to_human_readable({'bytes' => $space_needed})." (".$anvil->Convert->add_commas({number => $space_needed})." Bytes)] from the storage group: [".$storage_group_name."], but only: [".$anvil->Convert->bytes_to_human_readable({'bytes' => $space_on_dr})." (".$anvil->Convert->add_commas({number => $space_on_dr})." bytes)] is available on DR. Unable to proceed.\n";
|
||
|
|
$problem = 1;
|
||
|
|
}
|
||
|
|
}
|
||
|
|
if ($problem)
|
||
|
|
{
|
||
|
|
$anvil->nice_exit({exit_code => 1});
|
||
|
|
}
|
||
|
|
|
||
|
|
print "Verified that there is enough space on DR to proceed!\n";
|
||
|
|
print "The connection protocol will be: [".$anvil->data->{switches}{protocol}."]\n";
|
||
|
|
print "The following LV(s) will be created:\n";
|
||
|
|
foreach my $resource (sort {$a cmp $b} keys %{$anvil->data->{server}{drbd}})
|
||
|
|
{
|
||
|
|
foreach my $volume (sort {$a cmp $b} keys %{$anvil->data->{server}{drbd}{$resource}})
|
||
|
|
{
|
||
|
|
print "- Resource: [".$resource."], Volume: [".$volume."]\n";
|
||
|
|
my $lv_size = $anvil->data->{server}{drbd}{$resource}{$volume}{size};
|
||
|
|
my $storage_group_uuid = $anvil->data->{server}{drbd}{$resource}{$volume}{storage_group_uuid};
|
||
|
|
my $dr_lv_name = $resource."_".$volume;
|
||
|
|
my $dr_vg_name = $anvil->Storage->get_vg_name({
|
||
|
|
debug => 3,
|
||
|
|
storage_group_uuid => $storage_group_uuid,
|
||
|
|
host_uuid => $dr_host_uuid,
|
||
|
|
});
|
||
|
|
my $dr_lv_path = "/dev/".$dr_vg_name."/".$dr_lv_name;
|
||
|
|
my $extent_size = $anvil->data->{storage_groups}{storage_group_uuid}{$storage_group_uuid}{host_uuid}{$dr_host_uuid}{vg_extent_size};
|
||
|
|
my $extent_count = int($lv_size / $extent_size);
|
||
|
|
my $shell_call = $anvil->data->{path}{exe}{lvcreate}." -l ".$extent_count." -n ".$dr_lv_name." ".$dr_vg_name." -y";
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"s01:resource" => $resource,
|
||
|
|
"s02:volume" => $volume,
|
||
|
|
"s03:lv_size" => $anvil->Convert->add_commas({number => $lv_size})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $lv_size}).")", ,
|
||
|
|
"s04:storage_group_uuid" => $storage_group_uuid,
|
||
|
|
"s05:dr_lv_name" => $dr_lv_name,
|
||
|
|
"s06:dr_vg_name" => $dr_vg_name,
|
||
|
|
"s07:dr_lv_path" => $dr_lv_path,
|
||
|
|
"s08:extent_size" => $anvil->Convert->add_commas({number => $extent_size})." (".$anvil->Convert->bytes_to_human_readable({'bytes' => $extent_size}).")",
|
||
|
|
"s09:extent_count" => $extent_count,
|
||
|
|
"s10:shell_call" => $shell_call,
|
||
|
|
}});
|
||
|
|
|
||
|
|
$anvil->data->{server}{dr}{volumes}{$resource}{$volume}{lvcreate_call} = $shell_call;
|
||
|
|
$anvil->data->{server}{dr}{volumes}{$resource}{$volume}{lv_path} = $dr_lv_path;
|
||
|
|
$anvil->data->{server}{dr}{volumes}{$resource}{$volume}{storage_group_uuid} = $storage_group_uuid;
|
||
|
|
$anvil->data->{server}{dr}{volumes}{$resource}{$volume}{drbd_tcp_port} = $anvil->data->{server}{drbd}{$resource}{$volume}{tcp_port};
|
||
|
|
$anvil->data->{server}{dr}{volumes}{$resource}{$volume}{drbd_minor} = $anvil->data->{server}{drbd}{$resource}{$volume}{minor_number};
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"server::dr::volumes::${resource}::${volume}::lvcreate_call" => $anvil->data->{server}{dr}{volumes}{$resource}{$volume}{lvcreate_call},
|
||
|
|
"server::dr::volumes::${resource}::${volume}::lv_path" => $anvil->data->{server}{dr}{volumes}{$resource}{$volume}{lv_path},
|
||
|
|
"server::dr::volumes::${resource}::${volume}::storage_group_uuid" => $anvil->data->{server}{dr}{volumes}{$resource}{$volume}{storage_group_uuid},
|
||
|
|
"server::dr::volumes::${resource}::${volume}::drbd_tcp_port" => $anvil->data->{server}{dr}{volumes}{$resource}{$volume}{drbd_tcp_port},
|
||
|
|
"server::dr::volumes::${resource}::${volume}::drbd_minor" => $anvil->data->{server}{dr}{volumes}{$resource}{$volume}{drbd_minor},
|
||
|
|
}});
|
||
|
|
|
||
|
|
# Get the VG name that this volume will be created on.
|
||
|
|
print " - The LV: [".$dr_lv_path."] with the size: [".$anvil->Convert->bytes_to_human_readable({'bytes' => $lv_size})." (".$anvil->Convert->add_commas({number => $lv_size})." Bytes)] will be created.\n";
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
### NOTE: 'Yes' is set when a job is picked up, so this won't re-register the job.
|
||
|
|
my $record_job = 0;
|
||
|
|
if (not $anvil->data->{switches}{Yes})
|
||
|
|
{
|
||
|
|
# Ask the user to confirm.
|
||
|
|
print "\n- Proceed? [N/y]: ";
|
||
|
|
my $answer = <STDIN>;
|
||
|
|
chomp $answer;
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { answer => $answer }});
|
||
|
|
|
||
|
|
if ($answer =~ /^y/i)
|
||
|
|
{
|
||
|
|
print "- Thank you, storing job now.\n";
|
||
|
|
$record_job = 1;
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { record_job => $record_job }});
|
||
|
|
}
|
||
|
|
else
|
||
|
|
{
|
||
|
|
print "- Aborting.\n";
|
||
|
|
$anvil->nice_exit({exit_code => 0});
|
||
|
|
}
|
||
|
|
}
|
||
|
|
elsif (not $anvil->data->{switches}{'job-uuid'})
|
||
|
|
{
|
||
|
|
$record_job = 1;
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { record_job => $record_job }});
|
||
|
|
}
|
||
|
|
|
||
|
|
if ($record_job)
|
||
|
|
{
|
||
|
|
my $job_data = "server=".$anvil->data->{switches}{server}."\n";
|
||
|
|
$job_data .= "protect=1\n";
|
||
|
|
$job_data .= "protocol=".$anvil->data->{switches}{protocol}."\n";
|
||
|
|
|
||
|
|
# Register the job with this host
|
||
|
|
my ($job_uuid) = $anvil->Database->insert_or_update_jobs({
|
||
|
|
debug => 2,
|
||
|
|
job_command => $anvil->data->{path}{exe}{'anvil-manage-dr'}.$anvil->Log->switches,
|
||
|
|
job_data => $job_data,
|
||
|
|
job_name => "server::dr",
|
||
|
|
job_title => "job_0356",
|
||
|
|
job_description => "job_0357",
|
||
|
|
job_progress => 0,
|
||
|
|
job_host_uuid => $anvil->Get->host_uuid,
|
||
|
|
});
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { job_uuid => $job_uuid }});
|
||
|
|
$anvil->nice_exit({exit_code => 0});
|
||
|
|
}
|
||
|
|
|
||
|
|
return(0);
|
||
|
|
}
|
||
|
|
|
||
|
|
sub load_job
|
||
|
|
{
|
||
|
|
my ($anvil) = @_;
|
||
|
|
|
||
|
|
$anvil->Job->clear();
|
||
|
|
$anvil->Job->get_job_details();
|
||
|
|
$anvil->Job->update_progress({
|
||
|
|
progress => 1,
|
||
|
|
job_picked_up_by => $$,
|
||
|
|
job_picked_up_at => time,
|
||
|
|
message => "message_0263",
|
||
|
|
});
|
||
|
|
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"jobs::job_command" => $anvil->data->{jobs}{job_command},
|
||
|
|
"jobs::job_data" => $anvil->data->{jobs}{job_data},
|
||
|
|
"jobs::job_progress" => $anvil->data->{jobs}{job_progress},
|
||
|
|
"jobs::job_status" => $anvil->data->{jobs}{job_status},
|
||
|
|
}});
|
||
|
|
|
||
|
|
# Break up the job data into switches.
|
||
|
|
$anvil->data->{switches}{Yes} = 1;
|
||
|
|
foreach my $line (split/\n/, $anvil->data->{jobs}{job_data})
|
||
|
|
{
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { line => $line }});
|
||
|
|
if ($line =~ /(.*?)=(.*)$/)
|
||
|
|
{
|
||
|
|
my $key = $1;
|
||
|
|
my $value = $2;
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
key => $key,
|
||
|
|
value => $value,
|
||
|
|
}});
|
||
|
|
|
||
|
|
$anvil->data->{switches}{$key} = $value;
|
||
|
|
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
|
||
|
|
"switches::${key}" => $anvil->data->{switches}{$key},
|
||
|
|
}});
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
return(0);
|
||
|
|
}
|