[Pacemaker] node is unable to join cluster after upgrade (crmd dies)
Andrew Beekhof
andrew at beekhof.net
Tue Jun 17 23:13:56 UTC 2014
On 18 Jun 2014, at 2:59 am, Krause, Markus <krause at biochem.mpg.de> wrote:
> Hi all,
>
> I am using pacemaker/openais and drbd to have a high-available MySQL server which worked „for years“ (sind SLES 11 SP 0) without configuration chances.
> Just two days ago I did an update (zypper up) which now leads to the issue described below.
> Although the subject of my email is quite similar to many others on this list I could not find an equivalent (so it seems) problem or a solution. I tried to provide as much information as seemed to be reasonable to me so I apologize in advance for the many lines!
This looks ominous:
error: reap_dead_nodes: We're not part of the cluster anymore
It could be a software bug, but I'd start by checking the firewall.
Upgrades will sometimes affect those.
>
> Two servers with identical configuration (hard and software) running SLES 11 SP3 (with HAE) have the following crm config:
>
> ———————
> sql01a:~ # crm configure show
> node sql01a
> node sql01b
> primitive drbd_drbd0 ocf:linbit:drbd \
> params drbd_resource="drbd0" \
> op monitor interval="14" role="Master" \
> op monitor interval="16" role="Slave" \
> op start interval="0" timeout="240" \
> op stop interval="0" timeout="240"
> primitive ha_ip ocf:heartbeat:IPaddr2 \
> op monitor interval="5" timeout="20" \
> params ip="10.4.11.5" cidr_netmask="16" iflabel="ha" nic="eth0"
> primitive mount_drbd0 ocf:heartbeat:Filesystem \
> params device="/dev/drbd0" fstype="ext3" directory="/var/drbd0" \
> meta target-role="Started" \
> op start interval="0" timeout="60" \
> op stop interval="0" timeout="300"
> primitive mysqld ocf:heartbeat:mysql \
> params binary="/usr/sbin/mysqld" config="/etc/my.cnf" datadir="/var/drbd0/mysql/data" user="mysql" group="mysql" log="/var/log/mysql.log" pid="/var/lib/mysql/mysqld.pid" socket="/var/lib/mysql/mysql.sock" test_table="ha_check.mysql_check" test_user=„XXX" test_passwd=„XXX" enable_creation="no" \
> op monitor interval="10" timeout="60" \
> op start interval="0" timeout="240" \
> op stop interval="0" timeout="240"
> primitive ping_gateway ocf:pacemaker:ping \
> params dampen="10" multiplier="100" host_list="10.4.0.1" \
> op start interval="0" timeout="60" \
> op monitor interval="5" timeout="60"
> group group_mysqld mount_drbd0 mysqld ha_ip
> ms ms_drbd_drbd0 drbd_drbd0 \
> meta master-max="1" master-node-max="1" clone-max="2" clone-node-max="1" notify="true"
> clone ping_clone ping_gateway
> location loc_drbdmaster_drbd0 ms_drbd_drbd0 \
> rule $id="loc_drbdmaster_ping-rule" $role="Master" pingd: defined pingd
> colocation col_ms_drbd_drbd0_mount_drbd0 inf: group_mysqld ms_drbd_drbd0:Master
> order ord_ms_drbd_drbd0_mount_drbd0 inf: ms_drbd_drbd0:promote group_mysqld:start
> property $id="cib-bootstrap-options" \
> dc-version="1.1.10-f3eeaf4" \
> cluster-infrastructure="classic openais (with plugin)" \
> expected-quorum-votes="2" \
> no-quorum-policy="ignore" \
> stonith-enabled="false“
> ———————
>
> this worked since the first setup long ago.
> after the update (zypper up) two days ago only sql01b is running, sql01a is unable to join the cluster.
>
> ———————
> sql01a:~ # crm status:
> Last updated: Tue Jun 17 18:20:29 2014
> Last change: Tue Jun 17 17:26:20 2014 by root via crm_attribute on sql01b
> Stack: classic openais (with plugin)
> Current DC: sql01b - partition with quorum
> Version: 1.1.10-f3eeaf4
> 2 Nodes configured, 2 expected votes
> 7 Resources configured
>
>
> Online: [ sql01b ]
> OFFLINE: [ sql01a ]
>
> Clone Set: ping_clone [ping_gateway]
> Started: [ sql01b ]
> Stopped: [ sql01a ]
> Master/Slave Set: ms_drbd_drbd0 [drbd_drbd0]
> Masters: [ sql01b ]
> Stopped: [ sql01a ]
> Resource Group: group_mysqld
> mount_drbd0 (ocf::heartbeat:Filesystem): Started sql01b
> mysqld (ocf::heartbeat:mysql): Started sql01b
> ha_ip (ocf::heartbeat:IPaddr2): Started sql01b
> ———————
>
> Starting the cluster tools on sql01a („rcopenais start“) leads to the following lines (snipped the „OK-looking“ lines at top as I enabled debugging in corosync.conf; if necessary i can of course post the whole output, but that’s a lot of lines ...) in /var/log/messages:
>
> ———————
> [snip]
> Jun 17 18:20:18 sql01a corosync[14051]: [MAIN ] main.c:278 Completed service synchronization, ready to provide service.
> Jun 17 18:20:18 sql01a corosync[14051]: [TOTEM ] totempg.c:292 waiting_trans_ack changed to 0
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.cib failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.crmd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.cib failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.crmd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a mgmtd: [14062]: debug: main: run the loop...
> Jun 17 18:20:18 sql01a mgmtd: [14062]: info: Started.
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:486 ERROR: pcmk_wait_dispatch: Child process cib exited (pid=14056, rc=100)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:490 notice: pcmk_wait_dispatch: Child process cib no longer wishes to be respawned
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1672 debug: send_cluster_id: Local update: id=17499146, born=1452, seq=1452
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] utils.c:340 info: update_member: Node sql01a now has process list: 00000000000000000000000000151212 (1380882)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1672 debug: send_cluster_id: Local update: id=17499146, born=1452, seq=1452
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:486 ERROR: pcmk_wait_dispatch: Child process crmd exited (pid=14061, rc=100)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:490 notice: pcmk_wait_dispatch: Child process crmd no longer wishes to be respawned
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1672 debug: send_cluster_id: Local update: id=17499146, born=1452, seq=1452
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] utils.c:340 info: update_member: Node sql01a now has process list: 00000000000000000000000000151012 (1380370)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1672 debug: send_cluster_id: Local update: id=17499146, born=1452, seq=1452
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:486 ERROR: pcmk_wait_dispatch: Child process attrd exited (pid=14059, rc=100)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:490 notice: pcmk_wait_dispatch: Child process attrd no longer wishes to be respawned
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1672 debug: send_cluster_id: Local update: id=17499146, born=1452, seq=1452
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] utils.c:340 info: update_member: Node sql01a now has process list: 00000000000000000000000000150012 (1376274)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1672 debug: send_cluster_id: Local update: id=17499146, born=1452, seq=1452
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:486 ERROR: pcmk_wait_dispatch: Child process pengine exited (pid=14060, rc=100)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:490 notice: pcmk_wait_dispatch: Child process pengine no longer wishes to be respawned
> [snip]
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1365 info: send_member_notification: Sending membership update 1452 to 1 children
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:962 debug: pcmk_cluster_id_callback: Node update: sql01a (1.1.10)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:962 debug: pcmk_cluster_id_callback: Node update: sql01a (1.1.10)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.crmd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.crmd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.cib failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.crmd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.crmd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.attrd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.attrd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.attrd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:18 sql01a corosync[14051]: [pcmk ] plugin.c:1565 WARN: route_ais_message: Sending message to local.attrd failed: ipc delivery failed (rc=-2)
> Jun 17 18:20:22 sql01a sshd[14068]: Connection closed by 10.4.1.60 [preauth]
> Jun 17 18:20:23 sql01a corosync[14051]: [CONFDB] confdb.c:405 lib_init_fn: conn=0x663570
> Jun 17 18:20:23 sql01a corosync[14051]: [CONFDB] confdb.c:411 exit_fn for conn=0x663570
> Jun 17 18:20:27 sql01a stonith-ng[14057]: error: setup_cib: Could not connect to the CIB service: Transport endpoint is not connected (-107)
> Jun 17 18:20:27 sql01a stonith-ng[14057]: notice: crm_update_peer_state: plugin_handle_membership: Node sql01a[17499146] - state is now member (was (null))
> Jun 17 18:20:27 sql01a stonith-ng[14057]: notice: get_node_name: Could not obtain a node name for classic openais (with plugin) nodeid 34276362
> Jun 17 18:20:27 sql01a stonith-ng[14057]: notice: crm_update_peer_state: plugin_handle_membership: Node (null)[34276362] - state is now member (was (null))
> Jun 17 18:20:27 sql01a stonith-ng[14057]: notice: plugin_handle_membership: Membership 1452: quorum acquired
> [snip]
> ———————
>
> some more info:
> ———————
> sql01a:~ # corosync-cfgtool -s
> Printing ring status.
> Local node ID 17499146
> RING ID 0
> id = 10.4.11.1
> status = ring 0 active with no faults
> RING ID 1
> id = 10.255.4.1
> status = ring 1 active with no faults
> sql01a:~ # corosync-objctl | grep member
> runtime.totem.pg.mrp.srp.members.17499146.ip=r(0) ip(10.4.11.1) r(1) ip(10.255.4.1)
> runtime.totem.pg.mrp.srp.members.17499146.join_count=1
> runtime.totem.pg.mrp.srp.members.17499146.status=joined
> runtime.totem.pg.mrp.srp.members.34276362.ip=r(0) ip(10.4.11.2) r(1) ip(10.255.4.2)
> runtime.totem.pg.mrp.srp.members.34276362.join_count=1
> runtime.totem.pg.mrp.srp.members.34276362.status=joined
> ———————
> This is similar to the other (working) host.
>
> Interesting is:
> ———————
> sql01a:~ # rcopenais status
> Running
> sql01a:~ # ps ax | grep -E "pace|heart|coro" | grep -v grep
> 14051 ? Ssl 0:00 /usr/sbin/corosync
> 14057 ? S< 0:00 /usr/lib64/pacemaker/stonithd
> 14058 ? S< 0:00 /usr/lib64/pacemaker/lrmd
> 14062 ? S< 0:00 /usr/lib64/heartbeat/mgmtd
> sql01a:~ #
> ———————
>
>
> For digging further what’s going wrong I then started the missing processes (cib, attrd, pengine and crmd) from the command line (using the -V switch). cib, attrd and pengine are starting fine (even crm now shows a result, although „wrong“) but crmd fails:
>
> ———————
> sql01a:~ # /usr/lib64/pacemaker/crmd -V
> notice: main: CRM Git Version: f3eeaf4
> info: do_log: FSA: Input I_STARTUP from crmd_init() received in state S_STARTING
> info: config_find_next: Processing additional service options...
> info: get_config_opt: Found 'openais_clm' for option: name
> info: config_find_next: Processing additional service options...
> info: get_config_opt: Found 'openais_evt' for option: name
> info: config_find_next: Processing additional service options...
> info: get_config_opt: Found 'openais_ckpt' for option: name
> info: config_find_next: Processing additional service options...
> info: get_config_opt: Found 'openais_amf_v2' for option: name
> info: config_find_next: Processing additional service options...
> info: get_config_opt: Found 'openais_msg' for option: name
> info: config_find_next: Processing additional service options...
> info: get_config_opt: Found 'openais_lck' for option: name
> info: config_find_next: Processing additional service options...
> info: get_config_opt: Found 'openais_tmr' for option: name
> info: config_find_next: Processing additional service options...
> info: get_config_opt: Found 'pacemaker' for option: name
> info: get_config_opt: Found '0' for option: ver
> info: get_cluster_type: Detected an active 'classic openais (with plugin)' cluster
> info: do_cib_control: CIB connection established
> notice: crm_cluster_connect: Connecting to cluster infrastructure: classic openais (with plugin)
> info: init_cs_connection_classic: Creating connection to our Corosync plugin
> info: init_cs_connection_classic: AIS connection established
> notice: get_node_name: Defaulting to uname -n for the local classic openais (with plugin) node name
> info: plugin_get_details: Server details: id=17499146 uname=sql01a cname=pcmk
> info: crm_get_peer: Created entry 6002e998-b8ba-4daa-9c5d-1082bd1ab73a/0x65b8a0 for node sql01a/17499146 (1 total)
> info: crm_get_peer: Node 17499146 is now known as sql01a
> info: peer_update_callback: sql01a is now (null)
> info: crm_get_peer: Node 17499146 has uuid sql01a
> info: crm_update_peer_proc: init_cs_connection_classic: Node sql01a[17499146] - unknown is now online
> info: peer_update_callback: Client sql01a/peer now has status [online] (DC=<null>)
> info: init_cs_connection_once: Connection to 'classic openais (with plugin)': established
> notice: get_node_name: Defaulting to uname -n for the local classic openais (with plugin) node name
> info: do_ha_control: Connected to the cluster
> info: lrmd_ipc_connect: Connecting to lrmd
> info: do_lrm_control: LRM connection established
> info: do_started: Delaying start, no membership data (0000000000100000)
> info: do_started: Delaying start, no membership data (0000000000100000)
> info: do_started: Delaying start, no membership data (0000000000100000)
> notice: plugin_handle_membership: Membership 1460: quorum acquired
> info: crm_update_peer_proc: plugin_handle_membership: Node sql01a[17499146] - unknown is now member
> info: peer_update_callback: Client sql01a/peer now has status [offline] (DC=<null>)
> notice: peer_update_callback: Our peer connection failed
> notice: crm_update_peer_state: plugin_handle_membership: Node sql01a[17499146] - state is now member (was (null))
> info: peer_update_callback: sql01a is now member (was (null))
> info: crm_update_peer: plugin_handle_membership: Node sql01a: id=17499146 state=member addr=r(0) ip(10.4.11.1) r(1) ip(10.255.4.1) (new) votes=1 (new) born=1460 seen=1460 proc=00000000000000000000000000140012
> info: crm_get_peer: Created entry 8925fb86-2b70-4c91-af03-a8251a9182fc/0x65ecf0 for node sql01b/34276362 (2 total)
> info: crm_get_peer: Node 34276362 is now known as sql01b
> info: peer_update_callback: sql01b is now (null)
> info: crm_get_peer: Node 34276362 has uuid sql01b
> info: crm_update_peer_proc: plugin_handle_membership: Node sql01b[34276362] - unknown is now member
> info: peer_update_callback: Client sql01b/peer now has status [online] (DC=<null>)
> notice: crm_update_peer_state: plugin_handle_membership: Node sql01b[34276362] - state is now member (was (null))
> info: peer_update_callback: sql01b is now member (was (null))
> info: crm_update_peer: plugin_handle_membership: Node sql01b: id=34276362 state=member addr=r(0) ip(10.4.11.2) r(1) ip(10.255.4.2) (new) votes=1 (new) born=1436 seen=1460 proc=00000000000000000000000000151312
> error: reap_dead_nodes: We're not part of the cluster anymore
> info: plugin_handle_membership: Membership 1460: quorum retained
> error: do_log: FSA: Input I_ERROR from peer_update_callback() received in state S_STARTING
> notice: do_state_transition: State transition S_STARTING -> S_RECOVERY [ input=I_ERROR cause=C_CRMD_STATUS_CALLBACK origin=peer_update_callback ]
> warning: do_recover: Fast-tracking shutdown in response to errors
> error: do_started: Start cancelled... S_RECOVERY
> error: do_log: FSA: Input I_ERROR from reap_dead_nodes() received in state S_RECOVERY
> info: do_dc_release: DC role released
> info: do_te_control: Transitioner is now inactive
> error: do_log: FSA: Input I_TERMINATE from do_recover() received in state S_RECOVERY
> info: do_state_transition: State transition S_RECOVERY -> S_TERMINATE [ input=I_TERMINATE cause=C_FSA_INTERNAL origin=do_recover ]
> notice: lrm_state_verify_stopped: Stopped 0 recurring operations at shutdown (0 ops remaining)
> info: do_lrm_control: Disconnecting from the LRM
> info: lrmd_api_disconnect: Disconnecting from lrmd service
> info: lrmd_ipc_connection_destroy: IPC connection destroyed
> info: lrm_connection_destroy: LRM Connection disconnected
> info: lrmd_api_disconnect: Disconnecting from lrmd service
> notice: do_lrm_control: Disconnected from the LRM
> info: crm_cluster_disconnect: Disconnecting from cluster infrastructure: classic openais (with plugin)
> notice: terminate_cs_connection: Disconnecting from Corosync
> info: cluster_disconnect_cpg: No CPG connection
> info: crm_cluster_disconnect: Disconnected from classic openais (with plugin)
> info: do_ha_control: Disconnected from the cluster
> info: do_cib_control: Disconnecting CIB
> info: crmd_cib_connection_destroy: Connection to the CIB terminated...
> info: do_exit: Performing A_EXIT_0 - gracefully exiting the CRMd
> info: do_exit: [crmd] stopped (0)
> info: crmd_exit: Dropping I_RELEASE_SUCCESS: [ state=S_TERMINATE cause=C_FSA_INTERNAL origin=do_dc_release ]
> info: crmd_exit: Dropping I_TERMINATE: [ state=S_TERMINATE cause=C_FSA_INTERNAL origin=do_stop ]
> info: crmd_cs_destroy: connection closed
> info: crmd_init: 13510 stopped: OK (0)
> error: crmd_fast_exit: Could not recover from internal error
> info: crm_xml_cleanup: Cleaning up memory from libxml2
> sql01a:~ #
> ———————
>
> what I tried up to now:
> - stop sql01a, remove all cib.xml and crm.xml in their folders and restart sql01a (I know this is quite harsh but what did I have to loose …) -> the config on sql01a has been taken from sql01b but the crmd always dies as before (same error messages)
> - I removed all cluster related packages (pacemaker, corosync, drbd, …) from sql01a and deleted the remaining dirs and config files, reinstalled all with copies from sql01b -> same result.
> - I removed the whole crm config from sql01b (the working cluster host) using cibadmin -Q —force and re-added the config (using crm „commands") -> same result
> - of course rebooted/restarted both servers several times -> nothing changes …
>
>
> As deleting the config from sql01a (several times) I thought the error must be in the config/communication on/with sql01b but I also removed and added the config there, in addition there is now firewall between them and there is communication.
>
> Does anybody know/understand what is going on here?
> How can I solve this issue and get my cluster back to work as it did for quite a long time recently?
>
>
> thanks in advance for any hints!
>
> best regards,
> Markus
>
>
>
>
> -- .- .-. -.- ..- ... / -.- .-. .- ..- ... .
> Markus Krause
> Max Planck Institute of Biochemistry
> - Computing Center (core services, networking, OS X)
> Am Klopferspitz 18
> D-82152 Martinsried, Germany
> Phone: +49(89)8578-2825
>
> _______________________________________________
> 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
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 841 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20140618/e87da755/attachment-0004.sig>
More information about the Pacemaker
mailing list