[Pacemaker] hangs pending

Andrey Groshev greenx at yandex.ru
Fri Feb 14 04:37:28 UTC 2014


Yes, of course. Now beginning build world and test )

14.02.2014, 04:41, "Andrew Beekhof" <andrew at beekhof.net>:
> The previous patch wasn't quite right.
> Could you try this new one?
>
>    http://paste.fedoraproject.org/77123/13923376/
>
> [11:23 AM] beekhof at f19 ~/Development/sources/pacemaker/devel ☺ # git diff
> diff --git a/crmd/callbacks.c b/crmd/callbacks.c
> index ac4b905..d49525b 100644
> --- a/crmd/callbacks.c
> +++ b/crmd/callbacks.c
> @@ -199,8 +199,7 @@ peer_update_callback(enum crm_status_type type, crm_node_t * node, const void *d
>                  stop_te_timer(down->timer);
>
>                  flags |= node_update_join | node_update_expected;
> -                crm_update_peer_join(__FUNCTION__, node, crm_join_none);
> -                crm_update_peer_expected(__FUNCTION__, node, CRMD_JOINSTATE_DOWN);
> +                crmd_peer_down(node, FALSE);
>                  check_join_state(fsa_state, __FUNCTION__);
>
>                  update_graph(transition_graph, down);
> diff --git a/crmd/crmd_utils.h b/crmd/crmd_utils.h
> index bc472c2..1a2577a 100644
> --- a/crmd/crmd_utils.h
> +++ b/crmd/crmd_utils.h
> @@ -100,6 +100,7 @@ void crmd_join_phase_log(int level);
>  const char *get_timer_desc(fsa_timer_t * timer);
>  gboolean too_many_st_failures(void);
>  void st_fail_count_reset(const char * target);
> +void crmd_peer_down(crm_node_t *peer, bool full);
>
>  #  define fsa_register_cib_callback(id, flag, data, fn) do {              \
>          fsa_cib_conn->cmds->register_callback(                          \
> diff --git a/crmd/te_actions.c b/crmd/te_actions.c
> index f31d4ec..3bfce59 100644
> --- a/crmd/te_actions.c
> +++ b/crmd/te_actions.c
> @@ -80,11 +80,8 @@ send_stonith_update(crm_action_t * action, const char *target, const char *uuid)
>          crm_info("Recording uuid '%s' for node '%s'", uuid, target);
>          peer->uuid = strdup(uuid);
>      }
> -    crm_update_peer_proc(__FUNCTION__, peer, crm_proc_none, NULL);
> -    crm_update_peer_state(__FUNCTION__, peer, CRM_NODE_LOST, 0);
> -    crm_update_peer_expected(__FUNCTION__, peer, CRMD_JOINSTATE_DOWN);
> -    crm_update_peer_join(__FUNCTION__, peer, crm_join_none);
>
> +    crmd_peer_down(peer, TRUE);
>      node_state =
>          do_update_node_cib(peer,
>                             node_update_cluster | node_update_peer | node_update_join |
> diff --git a/crmd/te_utils.c b/crmd/te_utils.c
> index ad7e573..0c92e95 100644
> --- a/crmd/te_utils.c
> +++ b/crmd/te_utils.c
> @@ -247,10 +247,7 @@ tengine_stonith_notify(stonith_t * st, stonith_event_t * st_event)
>
>          }
>
> -        crm_update_peer_proc(__FUNCTION__, peer, crm_proc_none, NULL);
> -        crm_update_peer_state(__FUNCTION__, peer, CRM_NODE_LOST, 0);
> -        crm_update_peer_expected(__FUNCTION__, peer, CRMD_JOINSTATE_DOWN);
> -        crm_update_peer_join(__FUNCTION__, peer, crm_join_none);
> +        crmd_peer_down(peer, TRUE);
>       }
>  }
>
> diff --git a/crmd/utils.c b/crmd/utils.c
> index 3988cfe..2df53ab 100644
> --- a/crmd/utils.c
> +++ b/crmd/utils.c
> @@ -1077,3 +1077,13 @@ update_attrd_remote_node_removed(const char *host, const char *user_name)
>      crm_trace("telling attrd to clear attributes for remote host %s", host);
>      update_attrd_helper(host, NULL, NULL, user_name, TRUE, 'C');
>  }
> +
> +void crmd_peer_down(crm_node_t *peer, bool full)
> +{
> +    if(full && peer->state == NULL) {
> +        crm_update_peer_state(__FUNCTION__, peer, CRM_NODE_LOST, 0);
> +        crm_update_peer_proc(__FUNCTION__, peer, crm_proc_none, NULL);
> +    }
> +    crm_update_peer_join(__FUNCTION__, peer, crm_join_none);
> +    crm_update_peer_expected(__FUNCTION__, peer, CRMD_JOINSTATE_DOWN);
> +}
>
> On 16 Jan 2014, at 7:24 pm, Andrey Groshev <greenx at yandex.ru> wrote:
>
>>  16.01.2014, 01:30, "Andrew Beekhof" <andrew at beekhof.net>:
>>>  On 16 Jan 2014, at 12:41 am, Andrey Groshev <greenx at yandex.ru> wrote:
>>>>   15.01.2014, 02:53, "Andrew Beekhof" <andrew at beekhof.net>:
>>>>>   On 15 Jan 2014, at 12:15 am, Andrey Groshev <greenx at yandex.ru> wrote:
>>>>>>    14.01.2014, 10:00, "Andrey Groshev" <greenx at yandex.ru>:
>>>>>>>    14.01.2014, 07:47, "Andrew Beekhof" <andrew at beekhof.net>:
>>>>>>>>     Ok, here's what happens:
>>>>>>>>
>>>>>>>>     1. node2 is lost
>>>>>>>>     2. fencing of node2 starts
>>>>>>>>     3. node2 reboots (and cluster starts)
>>>>>>>>     4. node2 returns to the membership
>>>>>>>>     5. node2 is marked as a cluster member
>>>>>>>>     6. DC tries to bring it into the cluster, but needs to cancel the active transition first.
>>>>>>>>        Which is a problem since the node2 fencing operation is part of that
>>>>>>>>     7. node2 is in a transition (pending) state until fencing passes or fails
>>>>>>>>     8a. fencing fails: transition completes and the node joins the cluster
>>>>>>>>
>>>>>>>>     Thats in theory, except we automatically try again. Which isn't appropriate.
>>>>>>>>     This should be relatively easy to fix.
>>>>>>>>
>>>>>>>>     8b. fencing passes: the node is incorrectly marked as offline
>>>>>>>>
>>>>>>>>     This I have no idea how to fix yet.
>>>>>>>>
>>>>>>>>     On another note, it doesn't look like this agent works at all.
>>>>>>>>     The node has been back online for a long time and the agent is still timing out after 10 minutes.
>>>>>>>>     So "Once the script makes sure that the victim will rebooted and again available via ssh - it exit with 0." does not seem true.
>>>>>>>    Damn. Looks like you're right. At some time I broke my agent and had not noticed it. Who will understand.
>>>>>>    I repaired my agent - after send reboot he is wait STDIN.
>>>>>>    Returned "normally" a behavior - hangs "pending", until manually send reboot. :)
>>>>>   Right. Now you're in case 8b.
>>>>>
>>>>>   Can you try this patch:  http://paste.fedoraproject.org/68450/38973966
>>>>   Killed all day experiences.
>>>>   It turns out here that:
>>>>   1. Did cluster.
>>>>   2. On the node-2 send signal (-4) - killed corosink
>>>>   3. From node-1 (there DC) - stonith sent reboot
>>>>   4. Noda rebooted and resources start.
>>>>   5. Again. On the node-2 send signal (-4) - killed corosink
>>>>   6. Again. From node-1 (there DC) - stonith sent reboot
>>>>   7. Noda-2 rebooted and hangs in "pending"
>>>>   8. Waiting, waiting..... manually reboot.
>>>>   9. Noda-2 reboot and raised resources start.
>>>>   10. GOTO p.2
>>>  Logs?
>>  Yesterday I wrote an additional letter why not put the logs.
>>  Read it please, it contains a few more questions.
>>  Today again began to hang and continue along the same cycle.
>>  Logs here http://send2me.ru/crmrep2.tar.bz2
>>>>>>    New logs: http://send2me.ru/crmrep1.tar.bz2
>>>>>>>>     On 14 Jan 2014, at 1:19 pm, Andrew Beekhof <andrew at beekhof.net> wrote:
>>>>>>>>>      Apart from anything else, your timeout needs to be bigger:
>>>>>>>>>
>>>>>>>>>      Jan 13 12:21:36 [17223] dev-cluster2-node1.unix.tensor.ru stonith-ng: (  commands.c:1321  )   error: log_operation: Operation 'reboot' [11331] (call 2 from crmd.17227) for host 'dev-cluster2-node2.unix.tensor.ru' with device 'st1' returned: -62 (Timer expired)
>>>>>>>>>
>>>>>>>>>      On 14 Jan 2014, at 7:18 am, Andrew Beekhof <andrew at beekhof.net> wrote:
>>>>>>>>>>      On 13 Jan 2014, at 8:31 pm, Andrey Groshev <greenx at yandex.ru> wrote:
>>>>>>>>>>>      13.01.2014, 02:51, "Andrew Beekhof" <andrew at beekhof.net>:
>>>>>>>>>>>>      On 10 Jan 2014, at 9:55 pm, Andrey Groshev <greenx at yandex.ru> wrote:
>>>>>>>>>>>>>      10.01.2014, 14:31, "Andrey Groshev" <greenx at yandex.ru>:
>>>>>>>>>>>>>>      10.01.2014, 14:01, "Andrew Beekhof" <andrew at beekhof.net>:
>>>>>>>>>>>>>>>      On 10 Jan 2014, at 5:03 pm, Andrey Groshev <greenx at yandex.ru> wrote:
>>>>>>>>>>>>>>>>       10.01.2014, 05:29, "Andrew Beekhof" <andrew at beekhof.net>:
>>>>>>>>>>>>>>>>>        On 9 Jan 2014, at 11:11 pm, Andrey Groshev <greenx at yandex.ru> wrote:
>>>>>>>>>>>>>>>>>>         08.01.2014, 06:22, "Andrew Beekhof" <andrew at beekhof.net>:
>>>>>>>>>>>>>>>>>>>         On 29 Nov 2013, at 7:17 pm, Andrey Groshev <greenx at yandex.ru> wrote:
>>>>>>>>>>>>>>>>>>>>          Hi, ALL.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>          I'm still trying to cope with the fact that after the fence - node hangs in "pending".
>>>>>>>>>>>>>>>>>>>         Please define "pending".  Where did you see this?
>>>>>>>>>>>>>>>>>>         In crm_mon:
>>>>>>>>>>>>>>>>>>         ......
>>>>>>>>>>>>>>>>>>         Node dev-cluster2-node2 (172793105): pending
>>>>>>>>>>>>>>>>>>         ......
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>         The experiment was like this:
>>>>>>>>>>>>>>>>>>         Four nodes in cluster.
>>>>>>>>>>>>>>>>>>         On one of them kill corosync or pacemakerd (signal 4 or 6 oк 11).
>>>>>>>>>>>>>>>>>>         Thereafter, the remaining start it constantly reboot, under various pretexts, "softly whistling", "fly low", "not a cluster member!" ...
>>>>>>>>>>>>>>>>>>         Then in the log fell out "Too many failures ...."
>>>>>>>>>>>>>>>>>>         All this time in the status in crm_mon is "pending".
>>>>>>>>>>>>>>>>>>         Depending on the wind direction changed to "UNCLEAN"
>>>>>>>>>>>>>>>>>>         Much time has passed and I can not accurately describe the behavior...
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>         Now I am in the following state:
>>>>>>>>>>>>>>>>>>         I tried locate the problem. Came here with this.
>>>>>>>>>>>>>>>>>>         I set big value in property stonith-timeout="600s".
>>>>>>>>>>>>>>>>>>         And got the following behavior:
>>>>>>>>>>>>>>>>>>         1. pkill -4 corosync
>>>>>>>>>>>>>>>>>>         2. from node with DC call my fence agent "sshbykey"
>>>>>>>>>>>>>>>>>>         3. It sends reboot victim and waits until she comes to life again.
>>>>>>>>>>>>>>>>>        Hmmm.... what version of pacemaker?
>>>>>>>>>>>>>>>>>        This sounds like a timing issue that we fixed a while back
>>>>>>>>>>>>>>>>       Was a version 1.1.11 from December 3.
>>>>>>>>>>>>>>>>       Now try full update and retest.
>>>>>>>>>>>>>>>      That should be recent enough.  Can you create a crm_report the next time you reproduce?
>>>>>>>>>>>>>>      Of course yes. Little delay.... :)
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>      ......
>>>>>>>>>>>>>>      cc1: warnings being treated as errors
>>>>>>>>>>>>>>      upstart.c: In function ‘upstart_job_property’:
>>>>>>>>>>>>>>      upstart.c:264: error: implicit declaration of function ‘g_variant_lookup_value’
>>>>>>>>>>>>>>      upstart.c:264: error: nested extern declaration of ‘g_variant_lookup_value’
>>>>>>>>>>>>>>      upstart.c:264: error: assignment makes pointer from integer without a cast
>>>>>>>>>>>>>>      gmake[2]: *** [libcrmservice_la-upstart.lo] Error 1
>>>>>>>>>>>>>>      gmake[2]: Leaving directory `/root/ha/pacemaker/lib/services'
>>>>>>>>>>>>>>      make[1]: *** [all-recursive] Error 1
>>>>>>>>>>>>>>      make[1]: Leaving directory `/root/ha/pacemaker/lib'
>>>>>>>>>>>>>>      make: *** [core] Error 1
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>      I'm trying to solve this a problem.
>>>>>>>>>>>>>      Do not get solved quickly...
>>>>>>>>>>>>>
>>>>>>>>>>>>>      https://developer.gnome.org/glib/2.28/glib-GVariant.html#g-variant-lookup-value
>>>>>>>>>>>>>      g_variant_lookup_value () Since 2.28
>>>>>>>>>>>>>
>>>>>>>>>>>>>      # yum list installed glib2
>>>>>>>>>>>>>      Loaded plugins: fastestmirror, rhnplugin, security
>>>>>>>>>>>>>      This system is receiving updates from RHN Classic or Red Hat Satellite.
>>>>>>>>>>>>>      Loading mirror speeds from cached hostfile
>>>>>>>>>>>>>      Installed Packages
>>>>>>>>>>>>>      glib2.x86_64                                                              2.26.1-3.el6                                                               installed
>>>>>>>>>>>>>
>>>>>>>>>>>>>      # cat /etc/issue
>>>>>>>>>>>>>      CentOS release 6.5 (Final)
>>>>>>>>>>>>>      Kernel \r on an \m
>>>>>>>>>>>>      Can you try this patch?
>>>>>>>>>>>>      Upstart jobs wont work, but the code will compile
>>>>>>>>>>>>
>>>>>>>>>>>>      diff --git a/lib/services/upstart.c b/lib/services/upstart.c
>>>>>>>>>>>>      index 831e7cf..195c3a4 100644
>>>>>>>>>>>>      --- a/lib/services/upstart.c
>>>>>>>>>>>>      +++ b/lib/services/upstart.c
>>>>>>>>>>>>      @@ -231,12 +231,21 @@ upstart_job_exists(const char *name)
>>>>>>>>>>>>      static char *
>>>>>>>>>>>>      upstart_job_property(const char *obj, const gchar * iface, const char *name)
>>>>>>>>>>>>      {
>>>>>>>>>>>>      +    char *output = NULL;
>>>>>>>>>>>>      +
>>>>>>>>>>>>      +#if !GLIB_CHECK_VERSION(2,28,0)
>>>>>>>>>>>>      +    static bool err = TRUE;
>>>>>>>>>>>>      +
>>>>>>>>>>>>      +    if(err) {
>>>>>>>>>>>>      +        crm_err("This version of glib is too old to support upstart jobs");
>>>>>>>>>>>>      +        err = FALSE;
>>>>>>>>>>>>      +    }
>>>>>>>>>>>>      +#else
>>>>>>>>>>>>         GError *error = NULL;
>>>>>>>>>>>>         GDBusProxy *proxy;
>>>>>>>>>>>>         GVariant *asv = NULL;
>>>>>>>>>>>>         GVariant *value = NULL;
>>>>>>>>>>>>         GVariant *_ret = NULL;
>>>>>>>>>>>>      -    char *output = NULL;
>>>>>>>>>>>>
>>>>>>>>>>>>         crm_info("Calling GetAll on %s", obj);
>>>>>>>>>>>>         proxy = get_proxy(obj, BUS_PROPERTY_IFACE);
>>>>>>>>>>>>      @@ -272,6 +281,7 @@ upstart_job_property(const char *obj, const gchar * iface, const char *name)
>>>>>>>>>>>>
>>>>>>>>>>>>         g_object_unref(proxy);
>>>>>>>>>>>>         g_variant_unref(_ret);
>>>>>>>>>>>>      +#endif
>>>>>>>>>>>>         return output;
>>>>>>>>>>>>      }
>>>>>>>>>>>      Ok :) I patch source.
>>>>>>>>>>>      Type "make rc" - the same error.
>>>>>>>>>>      Because its not building your local changes
>>>>>>>>>>>      Make new copy via "fetch" - the same error.
>>>>>>>>>>>      It seems that if not exist ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3.tar.gz, then download it.
>>>>>>>>>>>      Otherwise use exist archive.
>>>>>>>>>>>      Cutted log .......
>>>>>>>>>>>
>>>>>>>>>>>      # make rc
>>>>>>>>>>>      make TAG=Pacemaker-1.1.11-rc3 rpm
>>>>>>>>>>>      make[1]: Entering directory `/root/ha/pacemaker'
>>>>>>>>>>>      rm -f pacemaker-dirty.tar.* pacemaker-tip.tar.* pacemaker-HEAD.tar.*
>>>>>>>>>>>      if [ ! -f ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3.tar.gz ]; then                                             \
>>>>>>>>>>>               rm -f pacemaker.tar.*;                                              \
>>>>>>>>>>>               if [ Pacemaker-1.1.11-rc3 = dirty ]; then                                   \
>>>>>>>>>>>                   git commit -m "DO-NOT-PUSH" -a;                                 \
>>>>>>>>>>>                   git archive --prefix=ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3/ HEAD | gzip > ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3.tar.gz;       \
>>>>>>>>>>>                   git reset --mixed HEAD^;                                        \
>>>>>>>>>>>               else                                                                \
>>>>>>>>>>>                   git archive --prefix=ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3/ Pacemaker-1.1.11-rc3 | gzip > ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3.tar.gz;     \
>>>>>>>>>>>               fi;                                                                 \
>>>>>>>>>>>               echo `date`: Rebuilt ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3.tar.gz;                                     \
>>>>>>>>>>>           else                                                                    \
>>>>>>>>>>>               echo `date`: Using existing tarball: ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3.tar.gz;                     \
>>>>>>>>>>>           fi
>>>>>>>>>>>      Mon Jan 13 13:23:21 MSK 2014: Using existing tarball: ClusterLabs-pacemaker-Pacemaker-1.1.11-rc3.tar.gz
>>>>>>>>>>>      .......
>>>>>>>>>>>
>>>>>>>>>>>      Well, "make rpm" - build rpms and I create cluster.
>>>>>>>>>>>      I spent the same tests and confirmed the behavior.
>>>>>>>>>>>      crm_reoprt log here - http://send2me.ru/crmrep.tar.bz2
>>>>>>>>>>      Thanks!
>>>>>>>>     ,
>>>>>>>>     _______________________________________________
>>>>>>>>     Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>>>>>>>>     http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>>>>>>
>>>>>>>>     Project Home: http://www.clusterlabs.org
>>>>>>>>     Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>>>>>>>     Bugs: http://bugs.clusterlabs.org
>>>>>>>    _______________________________________________
>>>>>>>    Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>>>>>>>    http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>>>>>
>>>>>>>    Project Home: http://www.clusterlabs.org
>>>>>>>    Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>>>>>>    Bugs: http://bugs.clusterlabs.org
>>>>>>    _______________________________________________
>>>>>>    Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>>>>>>    http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>>>>
>>>>>>    Project Home: http://www.clusterlabs.org
>>>>>>    Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>>>>>    Bugs: http://bugs.clusterlabs.org
>>>>>   ,
>>>>>   _______________________________________________
>>>>>   Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>>>>>   http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>>>
>>>>>   Project Home: http://www.clusterlabs.org
>>>>>   Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>>>>   Bugs: http://bugs.clusterlabs.org
>>>>   _______________________________________________
>>>>   Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>>>>   http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>>
>>>>   Project Home: http://www.clusterlabs.org
>>>>   Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>>>   Bugs: http://bugs.clusterlabs.org
>>>  ,
>>>  _______________________________________________
>>>  Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>>>  http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>
>>>  Project Home: http://www.clusterlabs.org
>>>  Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>>  Bugs: http://bugs.clusterlabs.org
>>  _______________________________________________
>>  Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>>  http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>
>>  Project Home: http://www.clusterlabs.org
>>  Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>  Bugs: http://bugs.clusterlabs.org
>
> ,
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org




More information about the Pacemaker mailing list