[ClusterLabs] ocf:heartbeart:apache and systemd

[Ken Gaillot]

On 03/20/2015 01:09 PM, Chuck Carmack wrote:
> Hi,
> 
> I'm on a RHEL7 system and noticing some problems starting and stopping 
> apache using systemctl vs. using pcs.
> The problems are with the $MAINPID env var used httpd.service.

Hi. Pacemaker and systemd are alternative means of managing services;
they shouldn't be used in combination. Any services managed by pacemaker
should be disabled in systemd, and only cluster commands (pcs etc.)
should be used to interact with them.

> Here are some examples:
> 
> 1) start apache with "pcs resoruce enable httpd"
> 
> 2) Attempt to stop apache with "systemctl stop httpd.service"
> --> nothing happens, apache is still running, no error from systemctl
> --> this must be because the MainPid used by httpd.service is not correct, 
> so the "kill WINCH ${MAINPID}" doesn't work
> 
> 3) Attempt to start apache using "systemctl start httpd.service"
> --> fails with:
> systemctl status httpd.service
> httpd.service - The Apache HTTP Server
>    Loaded: loaded (/usr/lib/systemd/system/httpd.service; enabled)
>    Active: failed (Result: exit-code) since Fri 2015-03-20 11:04:03 CDT; 
> 8s ago
>   Process: 25653 ExecStop=/bin/kill -WINCH ${MAINPID} (code=exited, 
> status=1/FAILURE)
>   Process: 25638 ExecStart=/usr/sbin/httpd $OPTIONS -DFOREGROUND 
> (code=exited, status=0/SUCCESS)
>  Main PID: 25638 (code=exited, status=0/SUCCESS)
> 
> Mar 20 11:04:03 vs404.rch.kstart.ibm.com httpd[25638]: httpd (pid 25202) 
> already running
> Mar 20 11:04:03 vs404.rch.kstart.ibm.com kill[25653]: kill: cannot find 
> process ""
> Mar 20 11:04:03 vs404.rch.kstart.ibm.com systemd[1]: httpd.service: 
> control process exited, code=exited status=1
> Mar 20 11:04:03 vs404.rch.kstart.ibm.com systemd[1]: Failed to start The 
> Apache HTTP Server.
> Mar 20 11:04:03 vs404.rch.kstart.ibm.com systemd[1]: Unit httpd.service 
> entered failed state.
> --> note, the same error happens on restart
> 
> --> Workaround:  I added a custom file for httpd.service containing these 
> lines:
> [Service]
> PIDFile=/var/run/httpd/httpd.pid
> --> did a "systemctl daemon-reload", and now "systemctl start 
> httpd.service" or restart works
> 
> 4) Now stop httpd using "systemctl stop httpd.service"
> --> The command works, and pacemaker restarts apache as it should, however 
> the MAINPID is now wrong,
> so we back to step (2)
> 
> Questions:
> -- Are there plans to better integrate ocf:heartbeat:apache with systemd? 
> Or are users expected to
> do the workaround as I did?
> -- Are there other integration issues with systemd?
> -- I'd like to workaround the problem in step (2) (stop does nothing).  I 
> think I can customize
> ExecStop in httpd.service but havne't figured out how.  Any advice how to 
> do this?
> 
> Thanks..
> Chuck
> 
> 
> 
> _______________________________________________
> Users mailing list: Users at clusterlabs.org
> http://clusterlabs.org/mailman/listinfo/users
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org
>