On Thu, Dec 30, 2010 at 10:24:12AM +0100, alexander.kra...@basf.com wrote:
> > On Wed, Dec 29, 2010 at 03:04:18PM +0100, Alexander Krauth wrote:
> > > # HG changeset patch
> > > # User Alexander Krauth <li...@sap.com>
> > > # Date 1293631454 -3600
> > > # Node ID a1f4bf0db5ff8c7c2ebd02e413df5e15201d4a7c
> > > # Parent 69cd9345a879e7764b4457834ded0093274d0322
> > > High: SAPInstance: Fixed monitor_clone function to ensure enqueue
> failover, in case of process (not host) failure
> > >
> > > RAs in versions <= 2.01 used a Heartbeat 2.0 specific feature to
> distinquish, if running in master or slave mode.
> > > This is not working with Pacemaker anymore.
> > >
> > > Since RA version 2.02 (not in official release) the monitor_clone
> function is damaged for the case of a local failure of the Standalone
> Enqueue process.
> > >
> > > This patch follows the requirement, that the RA must know be itself,
> if it is running in master or slave mode.
> > > Also it ensures, that always the salve (Enqueue Replication Server)
> gets promoted, if the master (Standalone Enqueue Server) fails.
> > >
> > > diff -r 69cd9345a879 -r a1f4bf0db5ff heartbeat/SAPInstance
> > > --- a/heartbeat/SAPInstance Wed Dec 29 14:40:41 2010 +0100
> > > +++ b/heartbeat/SAPInstance Wed Dec 29 15:04:14 2010 +0100
> > > @@ -32,6 +32,10 @@
> > > # OCF_RESKEY_PRE_STOP_USEREXIT (optional, lists a script which
> can be executed before the resource is stopped)
> > > # OCF_RESKEY_POST_STOP_USEREXIT (optional, lists a script which
> can be executed after the resource is stopped)
> > > #
> > > +# TODO: - Option to shutdown sapstartsrv for non-active instances ->
> that means: do probes only with OS tools (sapinstance_status)
> > > +# - Option for better standalone enqueue server monitoring,
> using ensmon (test enque-deque)
> > > +# - Option for cleanup abandoned enqueue replication tables
> > > +#
> > >
> #######################################################################
> > > # Initialization:
> > >
> > > @@ -68,7 +72,7 @@
> > > <?xml version="1.0"?>
> > > <!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
> > > <resource-agent name="SAPInstance">
> > > -<version>2.11</version>
> > > +<version>2.12</version>
> > >
> > > <shortdesc lang="en">Manages a SAP instance as an HA
> resource.</shortdesc>
> > > <longdesc lang="en">
> > > @@ -708,7 +712,7 @@
> > > #
> > > sapinstance_start_clone() {
> > > sapinstance_init $OCF_RESKEY_ERS_InstanceName
> > > - ${HA_SBIN_DIR}/crm_master -v 100 -l reboot
> > > + ${HA_SBIN_DIR}/crm_master -v 50 -l reboot
> > > sapinstance_start
> > > return $?
> > > }
> > > @@ -729,17 +733,38 @@
> > > # sapinstance_monitor_clone
> > > #
> > > sapinstance_monitor_clone() {
> > > - # Check status of potential master first
> > > + # first check with the status function (OS tools) if there could be
> something like a SAP instance running
> > > + # as we do not know here, if we are in master or slave state we do
> not want to start our monitoring
> > > + # agents (sapstartsrv) on the wrong host
> > > +
> > > sapinstance_init $OCF_RESKEY_InstanceName
> > > - sapinstance_monitor
> > > + sapinstance_status
> > > rc=$?
> > > - [ $rc -eq $OCF_SUCCESS ] && return $OCF_RUNNING_MASTER
> > > - [ $rc -ne $OCF_NOT_RUNNING ] && return $OCF_FAILED_MASTER
> > > -
> > > - # The master isn't running, and there were no errors, try ERS
> > > - sapinstance_init $OCF_RESKEY_ERS_InstanceName
> > > - sapinstance_monitor
> > > - rc=$?
> > > + if [ $rc -eq $OCF_SUCCESS ]; then
> > > + sapinstance_monitor
> > > + rc=$?
> > > + if [ $rc -eq $OCF_SUCCESS ]; then
> > > + ${HA_SBIN_DIR}/crm_master -Q -v 100 -l reboot
> > > + return $OCF_RUNNING_MASTER
> > > + else
> > > + ${HA_SBIN_DIR}/crm_master -v 10 -l reboot # by nature of
> the SAP enqueue server we have to make sure
> >
> > Shouldn't this be something like '-v -10'? I'm really not
> > sure, but if the master failed then this node may not be
> > capable of running the master.
>
> No, it should stay positive. This is for the rare case, that we do not
> have a slave running (slave failed, only one node active, ...).
> In that situation we want at least to do/try a local restart of the
> master.
> If there is a slave somewhere, it will have a higher value anyway.
>
> > > + # that we do a
> failover to the slave (enqueue replication server)
> > > + # in case the
> enqueue process has failed. We signal this to the
> > > + # cluster by
> setting our master preference to a lower value than the slave.
> > > + return $OCF_FAILED_MASTER
> > > + fi
> > > + else
> > > + sapinstance_init $OCF_RESKEY_ERS_InstanceName
> > > + sapinstance_status
> > > + rc=$?
> > > + if [ $rc -eq $OCF_SUCCESS ]; then
> > > + sapinstance_monitor
> > > + rc=$?
> > > + if [ $rc -eq $OCF_SUCCESS ]; then
> > > + ${HA_SBIN_DIR}/crm_master -Q -v 100 -l reboot
> > > + fi
> > > + fi
> > > + fi
> >
> > I got lost in this monitor function. A bit (hopefully) cleaner
> > version attached. Can you please review.
>
> Yes, I reviewed it. Looks very fine for me. Please apply your patch
> instead of mine.
Applied.
> Properly I can than do some testing next week with the final version.
OK.
Thanks,
Dejan
> > Thanks,
> > Dejan
>
> Regards,
> Alex
>
> > > return $rc
> > > }
> > > @@ -785,16 +810,25 @@
> > >
> > >
> > > #
> > > -# sapinstance_notify: After promotion of one master in the cluster,
> we make sure that all clones reset thier master
> > > -# value back to 100. This is because a failed
> monitor on a master might have degree one clone
> > > -# instance to score 10.
> > > +# sapinstance_notify: Handle master scoring - to make sure a slave
> gets the next master
> > > #
> > > sapinstance_notify() {
> > > local n_type="$OCF_RESKEY_CRM_meta_notify_type"
> > > local n_op="$OCF_RESKEY_CRM_meta_notify_operation"
> > >
> > > if [ "${n_type}_${n_op}" = "post_promote" ]; then
> > > + # After promotion of one master in the cluster, we make sure that
> all clones reset their master
> > > + # value back to 100. This is because a failed monitor on a master
> might have degree one clone
> > > + # instance to score 10.
> > > ${HA_SBIN_DIR}/crm_master -v 100 -l reboot
> > > + elif [ "${n_type}_${n_op}" = "pre_demote" ]; then
> > > + # if we are a slave and a demote event is anounced, make sure we
> have the highes wish to became master
> > > + # that is, when a slave resource was startet after the promote
> event of a already running master (e.g. node of slave was down)
> > > + # We also have to make sure to overrule the globaly set
> resource_stickiness or any fail-count factors => INFINITY
> > > + local n_uname="$OCF_RESKEY_CRM_meta_notify_demote_uname"
> > > + if [ ${n_uname} != ${HOSTNAME} ]; then
> > > + ${HA_SBIN_DIR}/crm_master -v INFINITY -l reboot
> > > + fi
> > > fi
> > > }
>
> _______________________________________________________
> Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
> Home Page: http://linux-ha.org/
_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/
