diff --git a/ocf/alteeve/server b/ocf/alteeve/server
index a1a89ad3..d8abcd1b 100755
--- a/ocf/alteeve/server
+++ b/ocf/alteeve/server
@@ -1239,50 +1239,14 @@ sub server_status
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, priority => "alert", key => "log_0331", variables => { timeout => $anvil->data->{environment}{OCF_RESKEY_CRM_meta_timeout} }});
}
- # Is 'libvirtd' running? We'll wait up to half the timeout for it to start (in case it _just_ started)
- # before timing out.
- my $wait_until = time + ($anvil->data->{environment}{OCF_RESKEY_CRM_meta_timeout} / 2000); # Devide by 2000 to convert to seconds and total second halved.
- my $look_for_pid = 0;
- my $libvirtd_wait = 1;
- my $warning_shown = 0;
- while($libvirtd_wait)
+ # libvirtd lists as disabled / stopped, but "starts" when called. So checking the daemon doesn't make
+ # sense. Given virsh might fail, if we don't find the server, we'll also look for it in 'ps'.
+ my $loop = 1;
+ while($loop)
{
- my $running = $anvil->System->check_daemon({daemon => "libvirtd.service"});
- $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { running => $running }});
- if ($running)
- {
- $libvirtd_wait = 0;
- $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { libvirtd_wait => $libvirtd_wait }});
- }
- else
- {
- # On EL8 and above, libvirtd starts on demand, so this error isn't
- if (not $warning_shown)
- {
- $anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 2, key => "log_0522", variables => { wait_time => ($wait_until - time) }});
- $warning_shown = 1;
- }
- sleep 1;
- if (time > $wait_until)
- {
- # Libvirtd isn't running, try to find the PID of the server (in case it's
- # running and libvirtd isn't)
- $anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 2, priority => "alert", key => "warning_0057"});
- $look_for_pid = 1;
- $libvirtd_wait = 0;
- $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
- look_for_pid => $look_for_pid,
- libvirtd_wait => $libvirtd_wait,
- }});
- }
- }
- }
-
- # If libvirtd wasn't running, we'll manually look for a PID.
- if ($look_for_pid)
- {
- my $server_up = 0;
- my $shell_call = $anvil->data->{path}{exe}{ps}." aux";
+ $loop = 0;
+ my $found = 0;
+ my $shell_call = $anvil->data->{path}{exe}{setsid}." --wait ".$anvil->data->{path}{exe}{virsh}." list --all";
$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});
@@ -1290,59 +1254,8 @@ sub server_status
output => $output,
return_code => $return_code,
}});
- foreach my $line (split/\n/, $output)
- {
- next if $line !~ /qemu-kvm/;
-
- $line = $anvil->Words->clean_spaces({ string => $line });
- $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { line => $line }});
-
- if ($line =~ /guest=(.*?),/)
- {
- my $this_server = $1;
- $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { this_server => $this_server }});
-
- if ($this_server eq $server)
- {
- # Found it.
- $server_up = 1;
- $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { server_up => $server_up }});
- last;
- }
- }
- }
-
- if ($server_up)
- {
- # The server is running. Exit with OCF_SUCCESS (rc 0);
- $anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "log_0523"});
- $anvil->nice_exit({exit_code => 0});
- }
- else
- {
- # The server is not running. Exit with OCF_NOT_RUNNING (rc: 7)
- $anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "log_0524"});
- $anvil->nice_exit({exit_code => 7});
- }
- }
- else
- {
- # Parse the virsh state. If it's listed as 'crashed', return OCF_ERR_GENERIC (rc: 1). If it's
- # 'in shutdown', 'loop' gets set to 1 and this will loop indefinitely. We don't put a timer
- # on it, we let pacemaker handle that.
- my $loop = 1;
- while($loop)
+ if (not $return_code)
{
- $loop = 0;
- my $found = 0;
- my $shell_call = $anvil->data->{path}{exe}{setsid}." --wait ".$anvil->data->{path}{exe}{virsh}." list --all";
- $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,
- }});
foreach my $line (split/\n/, $output)
{
$line = $anvil->Words->clean_spaces({ string => $line });
@@ -1351,7 +1264,7 @@ sub server_status
if ($line =~ /\s\Q$server\E\s+(.*)/)
{
my $state = $1;
- $found = 1;
+ $found = 1;
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => {
found => $found,
'state' => $state,
@@ -1382,7 +1295,6 @@ pmsuspended - The domain has been suspended by guest power management, e.g. ente
$anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { loop => $loop }});
sleep 1;
- last;
}
elsif ($state eq "shut off")
{
@@ -1401,11 +1313,53 @@ pmsuspended - The domain has been suspended by guest power management, e.g. ente
}
}
}
+ }
+
+ # If it wasn't found at all, exit.
+ if ((not $found) or ($return_code))
+ {
+ # If we've timed out, we'll look for it using it's PID.
+ my $server_up = 0;
+ my $shell_call = $anvil->data->{path}{exe}{ps}." aux";
+ $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,
+ }});
+ foreach my $line (split/\n/, $output)
+ {
+ next if $line !~ /qemu-kvm/;
+
+ $line = $anvil->Words->clean_spaces({ string => $line });
+ $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { line => $line }});
+
+ if ($line =~ /guest=(.*?),/)
+ {
+ my $this_server = $1;
+ $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { this_server => $this_server }});
+
+ if ($this_server eq $server)
+ {
+ # Found it.
+ $server_up = 1;
+ $anvil->Log->variables({source => $THIS_FILE, line => __LINE__, level => 2, list => { server_up => $server_up }});
+ last;
+ }
+ }
+ }
- # If it wasn't found at all, exit.
- if (not $found)
+ # Did we find it by PID?
+ if ($server_up)
+ {
+ # The server is running. Exit with OCF_SUCCESS (rc 0);
+ $anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "log_0523"});
+ $anvil->nice_exit({exit_code => 0});
+ }
+ else
{
- # Exit with OCF_NOT_RUNNING (rc: 7);
+ # The server is not running. Exit with OCF_NOT_RUNNING (rc: 7)
$anvil->Log->entry({source => $THIS_FILE, line => __LINE__, level => 1, key => "log_0526", variables => { server_name => $server }});
$anvil->nice_exit({exit_code => 7});
}
diff --git a/share/words.xml b/share/words.xml
index 1f333fb5..eee0a950 100644
--- a/share/words.xml
+++ b/share/words.xml
@@ -2377,10 +2377,10 @@ The file: [#!variable!file!#] needs to be updated. The difference is:
The server: [#!variable!server!#] is indeed running. It will be shut down now.
Checking the status of the server: [#!variable!server!#].
The 'libvirtd' daemon is not running. It may be starting up, will wait: [#!variable!wait_time!#] seconds...
- Found the server to be running using it's PID. The state of the server can't be determined, however. Please start the 'libvirtd' daemon!
+ Found the server to be running using it's PID. The state of the server can't be determined, however. There appears to be a problem with 'virsh'!
No PID for the server was found. It is not running on this host.
The server: [#!variable!server_name!#] is shutting down. Will wait for it to finish...
- The server: [#!variable!server_name!#] is off.
+ The server: [#!variable!server_name!#] is off (not found in virsh or by PID).
The server: [#!variable!server_name!#] is running (state is: [#!variable!state!#]).
We've been asked to migrating the server: [#!variable!server!#] to: [#!variable!target_host!#].
Checking server state after: [#!variable!server!#] was migrated to this host.