seasharp
/
anvil
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #241 from ClusterLabs/anvil-tools-dev

Browse Source 
* Changed the default trigger of live migrations to require a health …
main
digimer-bot 2 years ago committed by GitHub
parent c1a3f8099d 89121a2b3b
commit 4a7f9f79df
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
7 changed files with 72 additions and 28 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 10
      Anvil/Tools.pm
  2. 2
      Anvil/Tools/Alert.pm
  3. 4
      Anvil/Tools/Database.pm
  4. 14
      Anvil/Tools/ScanCore.pm
  5. 50
      scancore-agents/scan-apc-pdu/scan-apc-pdu
  6. 15
      share/words.xml
  7. 3
      tools/anvil-manage-files

10
Anvil/Tools.pm
Unescape View File

@ -1009,6 +1009,16 @@ sub _set_defaults
html => "alteeve",
},
};
$anvil->data->{feature} = {
scancore => {
disable => {
'preventative-live-migration' => 0,
},
threshold => {
'preventative-live-migration' => 2,
},
},
};
return(0);
}

2
Anvil/Tools/Alert.pm
Unescape View File

@ -276,7 +276,7 @@ sub check_condition_age
my $clear = defined $parameter->{clear} ? $parameter->{clear} : 0;
my $name = defined $parameter->{name} ? $parameter->{name} : "";
my $host_uuid = defined $parameter->{host_uuid} ? $parameter->{host_uuid} : "NULL";
my $host_uuid = defined $parameter->{host_uuid} ? $parameter->{host_uuid} : "";
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => {
clear => $clear,
name => $name,

4
Anvil/Tools/Database.pm
Unescape View File

@ -1625,7 +1625,7 @@ sub connect
# Read the DB identifier and then check that we've not already connected to this DB.
my $query = "SELECT system_identifier FROM pg_control_system();";
my $identifier = $anvil->Database->query({uuid => $uuid, query => $query, source => $THIS_FILE, line => __LINE__})->[0]->[0];
my $identifier = $anvil->Database->query({debug => $debug, uuid => $uuid, query => $query, source => $THIS_FILE, line => __LINE__})->[0]->[0];
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => {
query => $query,
identifier => $identifier,
@ -1691,7 +1691,7 @@ sub connect
variable_name => "database::".$uuid."::active",
});
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => { active_value => $active_value }});
if (not $active_value)
if ($active_value eq "0")
{
# If we're "retry", we just started up.
if (($retry) && ($is_local))

14
Anvil/Tools/ScanCore.pm
Unescape View File

@ -2022,13 +2022,25 @@ sub post_scan_analysis_node
# Last, evaluate health if we're otherwise OK
if ($peer_health > $local_health)
{
# The user may have set a migration threashold.
my $difference = $peer_health - $local_health;
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => { difference => $difference }});
if (not $anvil->data->{feature}{scancore}{threshold}{'preventative-live-migration'})
{
$anvil->data->{feature}{scancore}{threshold}{'preventative-live-migration'} = 2;
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => $debug, list => {
'feature::scancore::threshold::preventative-live-migration' => $anvil->data->{feature}{scancore}{threshold}{'preventative-live-migration'},
}});
}
# A user may disable health-based preventative live migrations.
if ($anvil->data->{feature}{scancore}{disable}{'preventative-live-migration'})
{
# Do nothing.
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 0, key => "message_0239"});
}
else
elsif ($difference >= $anvil->data->{feature}{scancore}{threshold}{'preventative-live-migration'})
{
# How long has this been the case?
my $age = $anvil->Alert->check_condition_age({

50
scancore-agents/scan-apc-pdu/scan-apc-pdu
Unescape View File

@ -260,7 +260,7 @@ FROM
my $scan_apc_pdu_link_speed = $row->[10];
my $scan_apc_pdu_phase_count = $row->[11];
my $scan_apc_pdu_outlet_count = $row->[12];
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 3, list => {
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
scan_apc_pdu_uuid => $scan_apc_pdu_uuid,
scan_apc_pdu_fence_uuid => $scan_apc_pdu_fence_uuid,
scan_apc_pdu_serial_number => $scan_apc_pdu_serial_number,
@ -1385,6 +1385,7 @@ WHERE
}
# Delete this from the SQL hash so we know it didn't vanish.
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 2, key => "log_0726", variables => { hash_key => "sql::scan_apc_pdu_uuid::${scan_apc_pdu_uuid}" }});
delete $anvil->data->{sql}{scan_apc_pdu_uuid}{$scan_apc_pdu_uuid};
}
else
@ -1596,10 +1597,19 @@ INSERT INTO
scan_apc_pdu_ipv4_address => $scan_apc_pdu_ipv4_address,
}});
if ($scan_apc_pdu_model_number ne "DELETED")
{
# Yup! send an alert.
my $query = "
# The PDUs only allow one connection at a time, so if another scan agent is
# connected, we'll get this issue. As such, check how long it's been missing, and
# alert only if it's been missing for 10 minutes.
my $age = $anvil->Alert->check_condition_age({name => "scan_apc_pdu::lost_pdu::".$scan_apc_pdu_serial_number});
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { age => $age }});
if ($age > 600)
{
# Yup! send an alert.
my $query = "
UPDATE
scan_apc_pdus
SET
@ -1608,22 +1618,23 @@ SET
WHERE
scan_apc_pdu_uuid = ".$anvil->Database->quote($scan_apc_pdu_uuid)."
;";
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }});
push @{$anvil->data->{sys}{queries}}, $query;
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { query => $query }});
push @{$anvil->data->{sys}{queries}}, $query;
my $variables = {
model => $scan_apc_pdu_model_number,
serial_numer => $scan_apc_pdu_serial_number,
ip_address => $scan_apc_pdu_ipv4_address,
};
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_apc_pdu_message_0040", variables => $variables});
$anvil->Alert->register({
alert_level => "warning",
message => "scan_apc_pdu_message_0040",
variables => $variables,
set_by => $THIS_FILE,
sort_position => $anvil->data->{'scan-apc-pdu'}{alert_sort}++,
});
my $variables = {
model => $scan_apc_pdu_model_number,
serial_numer => $scan_apc_pdu_serial_number,
ip_address => $scan_apc_pdu_ipv4_address,
};
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "scan_apc_pdu_message_0040", variables => $variables});
$anvil->Alert->register({
alert_level => "warning",
message => "scan_apc_pdu_message_0040",
variables => $variables,
set_by => $THIS_FILE,
sort_position => $anvil->data->{'scan-apc-pdu'}{alert_sort}++,
});
}
}
}
@ -2004,6 +2015,9 @@ sub gather_pdu_data
# If I got the serial number, I found the PDU.
next if not $scan_apc_pdu_serial_number;
# In case this PDU disappeared before, this will clear that condition.
$anvil->Alert->check_condition_age({clear => 1, name => "scan_apc_pdu::lost_pdu::".$scan_apc_pdu_serial_number});
#############################################################################################
# Base PDU info #
#############################################################################################

15
share/words.xml
Unescape View File

@ -466,7 +466,7 @@ Giving up.
<key name="error_0341">Failed to find the server: [#!variable!server!#] by name or UUID? Exiting.</key>
<key name="error_0342">The protocol: [#!variable!protocol!#] is invalid. Please use '--help' for more information.</key>
<key name="error_0343">The DR host: [#!variable!host_name!#] doesn't appear to be storage group: [#!variable!storage_group!#]. Unable to proceed.</key>
<key name="error_0344">We need: [#!variable!space_needed!# (#!variables!space_needed_bytes!# Bytes)] from the storage group: [#!variable!storage_group!#], but only: [#!variables!space_on_dr!# (#!variable!space_on_dr_bytes!# bytes)] is available on DR. Unable to proceed.</key>
<key name="error_0344">We need: [#!variable!space_needed!# (#!variable!space_needed_bytes!# Bytes)] from the storage group: [#!variable!storage_group!#], but only: [#!variable!space_on_dr!# (#!variable!space_on_dr_bytes!# bytes)] is available on DR. Unable to proceed.</key>
<key name="error_0345">[ Error ] - The check appears to have failed. Expected a return code of '0', but got: [#!variable!return_code!#]
The output, if any, was
====
@ -520,7 +520,7 @@ The definition data passed in was:
====
]]></key>
<key name="error_0368">[ Error ] - Failed to wipe and delete the logical volume: [#!variable!local_lv!#] that was volume number: [#!variable!volume!#] under the server: [#!variable!server!#].</key>
<key name="error_0369">There was a problem deleting: [#!variables!config_file!#]. The rest of the process completed successfully. Please manually remove this file if it still exists.</key>
<key name="error_0369">There was a problem deleting: [#!variable!config_file!#]. The rest of the process completed successfully. Please manually remove this file if it still exists.</key>
<!-- Files templates -->
<!-- NOTE: Translating these files requires an understanding of which lines are translatable -->
@ -576,7 +576,13 @@ sys::privacy::strong = #!data!sys::privacy::strong!#
# Normally, if one node in the Anvil! is healthier than the other, it will pull the servers from the peer
# on to it. This is a process called "preventative live migration". If you would like to disable this
# feature, set this to '1'.
feature::scancore::disable::preventative-live-migration = 0
#feature::scancore::disable::preventative-live-migration = 0
# If "preventative live migration" is enabled, this sets the threshold to trigger migration. The difference
# in health score has to be equal to or greater than the number below. The health scores are usually set to
# '1' per event, though scan agents are free to assign higher scores per event. The default threshold is
# '2'. To migrate on any health difference, set this to '1'. Use whole numbers only.
#feature::scancore::threshold::preventative-live-migration = 2
### Database
# Database connections;
@ -2191,8 +2197,9 @@ The file: [#!variable!file!#] needs to be updated. The difference is:
<key name="log_0721">The server: [#!variable!server!#] is ready to boot.</key>
<key name="log_0722">The server: [#!variable!server!#] was found to be running already, but it wasn't marked as booted. Marking it as if it just booted to handle any dependent servers.</key>
<key name="log_0723">The server: [#!variable!server!#] is configured to stay off, ignoring it.</key>
<key name="log_0724">The file: [#!variable!file!#] needs to be added to the database, but since the last scan it's size grew from: [#!variable!old_size_bytes!# (#!variables!old_size_hr!#)] to: [#!variable!new_size_bytes!# (#!variables!new_size_hr!#)]. A difference of: [#!variable!difference_bytes!# (#!variables!difference_hr!#)]. It might still be being uploaded, so we'll keep checking periodocally until the size stops changing.</key>
<key name="log_0724">The file: [#!variable!file!#] needs to be added to the database, but since the last scan it's size grew from: [#!variable!old_size_bytes!# (#!variable!old_size_hr!#)] to: [#!variable!new_size_bytes!# (#!variable!new_size_hr!#)]. A difference of: [#!variable!difference_bytes!# (#!variable!difference_hr!#)]. It might still be being uploaded, so we'll keep checking periodocally until the size stops changing.</key>
<key name="log_0725">Found the missing file: [#!variable!file!#] in the directory: [#!variable!directory!#]. Updating the database now.</key>
<key name="log_0726">Deleting the hash key: [#!variable!hash_key!#].</key>
<!-- Messages for users (less technical than log entries), though sometimes used for logs, too. -->
<key name="message_0001">The host name: [#!variable!target!#] does not resolve to an IP address.</key>

3
tools/anvil-manage-files
Unescape View File

@ -28,7 +28,8 @@
# 6 = The file to delete is not under '/mnt/shared/'.
#
# TODO:
# -
# - If two Strikers have the same file name, but different sizes, we get into a yo-yo of updating the two
# sides. If this happens, we need to rsync the larger one over the smaller one.
#
# NOTE:
# - remove unsyncs, add syncs.

Write Preview
Loading…
Cancel
Save