Newbie questions - 2008R2 2-Node Cluster and Failures

-

trying troubleshot sequence of events of outage on 2-node 2008 r2 mscs based cluster (we have ip address , sql server instance clustered).  refer nodes node05 , node06.

both nodes running on vmware esxi 5.x database , quorum disks attached via vmware rdm ibm xiv via fiber channel.  node06's rdm set use fixed-path addressing while node05's rdm (incorrectly) set use round-robin multipath (working correct this).  each node running on different blade center.  there private heartbeat network.

at beginning of event, node05 primary.

at 22:08:20, both nodes report node05 removed active failover cluster membership.  23 seconds later, both nodes report disks 'unexpectedly lost' respective node.  these errors continue on more 2 minutes before things seem come online.

our investigation shows that, @ least esx perspective, connectivity san luns not lost.  san monitoring shows nothing "dropping".  in addition, i'm not seeing in os/system event logs indicating storage lost -- disk errors show in cluster logs.  i'm not thinking sort of san disruption trigger event, want ensure theory fits how 2-node mscs cluster functions.

i'm theorizing node remove event (possibly triggered network disruption) occurred first may have triggered scsi-3 based "fencing" have resulted in disks appearing unavailable on both nodes though san still up.  however, understanding scsi reservation requests subsequent scsi reset occurs in "split" happens @ staggered intervals (three seconds "primary" node , 7 seconds "challenger" nodes) should resolved -- not 2+ minutes saw.

can confirm i'm on right track thinking?  or possibly describe how typical failure scenario play out if heartbeat network disrupted period of time?

first, configuration pass cluster storage validation tests? if not, needs sorted before doing else.

next, node05 being removed active failover cluster membership indication there sort of network connectivity issue between nodes. whatever reason, node06 lost connectivity node05, , assume node06 owns witness resource , able maintain quorum, causes node05 drop out of cluster.

when node05 drops out of cluster, other cluster node forcefully attempts take on ownership of resources owned node05. cause event logs on node05 filled errors reservation lost , disk errors regarding host being unable flush cache disk. expected in situation.

you shouldn't see same loss of reservation messages on node06. not expected behavior. i'd take @ cluster logs see might have happened while node attempting bring disks online. 

visit blog multi-site clustering



Windows Server  >  High Availability (Clustering)



Comments

Post a Comment

Popular posts from this blog

2008 Windows Deployment Server Properties Error

-
currently wading through setup , configuration of wds on 2008 server. have captured xp sp2 image , purposes of testing not prestaging. pxe response settings have been set "respond (known & unknown) client computers" leaving "for unknown clients, notify administrator etc" unchecked when ok properties sheet error "server properties error" "the system cannot find file specified" when attempting use command "wdsutil /approve-autoadddevices /requestid:1" in response pending devices within gui "an error occurred while trying execute command. error code: 0x6 error description: access denied" surely if not attempting add workstations domain (no prestaging) there shouldn't problem. any appreciated  answering part of own question although if can explain reason appreciated. server properties error occurs when launching windows deployment services using start>all programs>administrative tools. if access via server manager...
Read more

Can no longer user MS Update - Files required to use Microsoft Update are no longer registered

-
hi, i'm new server 2003.  installed it.  removed enhanced ie security.  proceed download , install windows server 2003 sp1, of dot net (1.1, 2.0, 3.0, 3.5, , updates), , sp2. after downloaded , installed of above, tried windows update menu option in ie.  (ie 8 windows server 2003) , sits there "done" in lower left hand corner.  never brings in window. so try www.microsoft.com -> download , trails -> update.  received error message "files required use microsoft update no longer registered or installed on computer".  click on register or reinstall files me now.  window changes , "check latest version of windows updating software .... ".   sits there (last time sat there on hour). i went on internet , found instructions: regsvr32 wuapi.dll regsvr32 wups.dll regsvr32 wuaueng.dll regsvr32 wucltui.dll regsvr32 atl.dll regsvr32 msxml3.dll did not work either.  have other suggestions?...
Read more

How do a find data in one file, search for it in another file and if not found, write a custom message to another file

-
in file if.txt, starting on 2nd line of file, data formatted follows: ah10551,a10551,cn=a10551,a10551@qts.com ah10600,a10600,cn=a10600,a10600@qts.com ah10608,a10608,cn=a10608,a10608@qts.com in file sf.txt, data formatted follows: 2011-02-16 17:26:44 sid acme\ah10551 added sid history of qts\a10551 2011-02-16 17:26:44 sid acme\ah10244 added sid history of qts\a10244 2011-02-16 17:26:45 sid acme\ah10425 added sid history of qts\a10425 please provide suggestions on how following: 1. in if.txt file, read first field first ',' , store in variable called $if1 2. in sf.txt file, search for the value stored in $if1 , if found, ignore.  if not found, write custom message in another file such as: $if1 sidhistory not migrated. these files wont large.  im trying build 'exception file' of items in if.txt file not occurring in sf.txt file. thanks in advance ideas. mtv99 many ways that... here's one... (using example data) [array]$if1=@() gc d:\if.txt...
Read more