Revision 51f90897
Added by Alexis Mousset over 7 years ago
| techniques/system/common/1.0/promises.st | ||
|---|---|---|
|
{
|
||
|
vars:
|
||
|
|
||
|
windows::
|
||
|
# process_term defines how many maximum instances of this
|
||
|
# binary should be running before attempting to SIGTERM them.
|
||
|
# process_kill is the same for SIGKILL.
|
||
|
!windows::
|
||
|
# On windows, cf-execd is a service, and there can be only one instance of it running (by design)
|
||
|
"process_term[execd]" string => "2";
|
||
|
"process_kill[execd]" string => "5";
|
||
|
|
||
|
any::
|
||
|
"process_term[agent]" string => "5";
|
||
|
"process_kill[agent]" string => "8";
|
||
|
|
||
| ... | ... | |
|
"pass1" expression => "any";
|
||
|
|
||
|
methods:
|
||
|
pass3::
|
||
|
pass3.!windows::
|
||
|
|
||
|
"any" usebundle => rudder_common_report("Common", "result_na", "&TRACKINGKEY&", "Process checking", "None", "CFEngine proccesses check is done by the rudder-agent CRON job");
|
||
|
|
||
|
pass3.windows::
|
||
|
"any" usebundle => rudder_common_report("Common", "result_success", "&TRACKINGKEY&", "Process checking", "None", "There is an acceptable number of CFEngine processes running on the machine"),
|
||
|
# Here, I can not use the binaries variable as CFEngine will iterate and output two reports, breaking the reporting.
|
||
|
ifvarclass => "!agent_has_gone_wild.!agent_has_gone_really_wild.!execd_has_gone_wild.!execd_has_gone_really_wild";
|
||
| ... | ... | |
|
ifvarclass => "${binaries}_has_gone_really_wild";
|
||
|
|
||
|
processes:
|
||
|
!windows::
|
||
|
"${sys.workdir}/bin/cf-serverd" restart_class => "start_server";
|
||
|
"${sys.workdir}/bin/cf-execd" restart_class => "start_executor";
|
||
|
|
||
|
# If there are more than 2 cf-execd's, it means cf-execd is starting to
|
||
|
# go crazy, so we ask politely to these processes to shut down.
|
||
|
|
||
|
"${sys.workdir}/bin/cf-${binaries}"
|
||
|
process_count => check_range("${binaries}", "0","${process_term[${binaries}]}"),
|
||
|
signals => { "term" },
|
||
|
classes => if_repaired("${binaries}_has_gone_wild"),
|
||
|
comment => "Checking if cf-${binaries} has gone wild";
|
||
|
|
||
|
# If there are too much cf-execd's/cf-agents running, it means that they are really
|
||
|
# going crazy. Let's be a bit less polite and more violent about killing them.
|
||
|
#
|
||
|
# These two promises overlap, because when you go past the 2/5-limit treshold,
|
||
|
# you still leave a chance for them to die with SIGTERM before the SIGKILL.
|
||
|
#
|
||
|
# Reason: The backend databases that stores the classes and some runtime
|
||
|
# parameters do really not appreciate beeing killed violently and may prevent
|
||
|
# the agent from operating properly.
|
||
|
|
||
|
"${sys.workdir}/bin/cf-${binaries}"
|
||
|
process_count => check_range("${binaries}", "0","${process_kill[${binaries}]}"),
|
||
|
signals => { "kill" },
|
||
|
classes => if_repaired("${binaries}_has_gone_really_wild"),
|
||
|
comment => "Checking if cf-${binaries} has gone really wild";
|
||
|
|
||
|
!windows.should_disable_server::
|
||
|
"${sys.workdir}/bin/cf-serverd"
|
||
|
signals => { "term", "kill" },
|
||
|
action => immediate;
|
||
|
|
||
|
windows::
|
||
|
# Using the path on windows fails, as process are not reported the same way
|
||
|
# And unfortunately, the cf-serverd is not a service.
|
||
| ... | ... | |
|
action => u_ifwin_bg,
|
||
|
classes => outcome("server");
|
||
|
|
||
|
start_executor::
|
||
|
"${sys.cf_execd}"
|
||
|
action => u_ifwin_bg,
|
||
|
classes => outcome("executor");
|
||
|
|
||
|
}
|
||
|
|
||
|
|
||
Also available in: Unified diff
Fixes #7381: Process management issues on nodes hosting LXC containers