----- Original Message ----- > From: "Ulrich Windl" <[email protected]> > To: "General Linux-HA mailing list" <[email protected]> > Sent: Friday, June 14, 2013 1:34:58 AM > Subject: [Linux-HA] Antw: ocf HA_RSCTMP directory location > > Hi! > > I think the location of the temporary directory is not that important,
It is because pacemaker has to make sure the directory exists every time it starts up. Otherwise agents will fail. > because you don't exchange data between RAs. For the design, I think it's > sufficient if an RA that actually uses the temporary directory does check > ist existence (for "validate-all", maybe). But even that is not necessary, > because if the RA cannot write it's PID where it wanted to, the star > toperation should fail, and the user will notice the problem. > > Amazingly I had seen a situation with samba recently, where smbd start exited > OK, but did not start, because the PID file (an obsolete one) already > existed. I had to remove the PID file manually. Clearly a bug in samba. If the temp directories don't get cleaned out on restart, this is a possibility. > > I think RAs should not rely on the fact that temp directories are clean when > a resource is going to be started. The resource tmp directory has to get cleaned out on startup, if it doesn't I don't think there is a good solution for resource agents to detect a stale pid file from one that is current. Nearly all the agents depend on this tmp directory to get reinitialized. If we decided not to depend on this logic, every agent would have to be altered to account for this. This would mean adding a layer of complexity to the agents that should otherwise be unnecessary. -- Vossel > > Regards, > Ulrich > > >>> David Vossel <[email protected]> schrieb am 13.06.2013 um 23:59 in > >>> Nachricht > <[email protected]>: > > Hey, > > > > Andrew and I have been running into some inconsistencies between > > resource-agent packages that we need to get cleared up. > > > > There's an ocf variable, HA_RSCTMP, used in many of the resource agents > > that > > represents a place the agent's can store their PID files and other > > temporary > > data. This data needs to live under some directory in /var/run as that > > directory is typically cleared on startup. This is important to prevent > > stale PID files and other transient data from being persistent across > > restarts. > > > > Anyway. Here's the problem. > > > > Pacemaker thinks that data should live in '/var/run/heartbeat/rsctmp', but > > not all the resource-agent packages are consistent with that. For example, > > Suse's resource-agent package sets HA_RSCTMP to '/var/run/resource-agents' > > ( > > looking at this rpm, > > http://download.opensuse.org/distribution/11.4/repo/oss/suse/x86_64/resource-a > > gents-1.0.3-9.12.1.x86_64.rpm ) > > > > We need to come to some sort of agreement because ultimately Pacemaker > > needs > > to make sure this directory exists on startup, whatever it is. If > > pacemaker > > doesn't create the right directory, it's possible the resource agents won't > > be able to access it since /var/run is re-initialized on startup. > > > > so, > > HA_RSCTMP = /var/run/heartbeat/rsctmp > > or > > HA_RSCTMP = /var/run/resource-agents > > > > thoughts? > > > > -- Vossel > > > > _______________________________________________ > > Linux-HA mailing list > > [email protected] > > http://lists.linux-ha.org/mailman/listinfo/linux-ha > > See also: http://linux-ha.org/ReportingProblems > > > > _______________________________________________ > Linux-HA mailing list > [email protected] > http://lists.linux-ha.org/mailman/listinfo/linux-ha > See also: http://linux-ha.org/ReportingProblems > _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
