[
https://issues.apache.org/jira/browse/HBASE-6060?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13291562#comment-13291562
]
ramkrishna.s.vasudevan commented on HBASE-6060:
-----------------------------------------------
bq.a Map of dead servers to their current set of region plans and
regions-in-transition.
We are only maintaining regions and the RITS. not the regionplans
bq.Does this sound right?
Yes. Exactly. All this decision made on the current regionPlan.
bq.Is this patch even handling the root cause of this JIRA: i.e. dealing w/
OPENING znodes that were made against the dead server?
Yes.
bq.In AM#processServerShutdown, we iterate the list of regions we get back from
AM#this.servers and from this set we will remove RIT. But won't this original
set of regions be missing regions that are OPENING or even OPENED but not yet
handled because its only after the OPENED region has updated its znode and the
znode has been removed by the OPEN handler do we add a region to
AM#this.servers? Or is the OPENING region handled elsewhere?
To answer this question, for the first part. If region R1 was initially in RS
A and now it is getting moved to RS B. If the RS A is going down then the
existing SSH flow for RS A will get the regions from AM#this.servers and remove
the same from this.regions. Because already something is in RIT it will not
assign it and so assignment happens smoothly to RS B.
Now take a case where the RS B goes down. In the existing code no where we
assigned those regions and left to the TM to assign. Now that is where this
patch comes into place and tries to assign it based on the regionplan. When
SSH sees that the dead server is in destination of the regionplan it will add
to the new data structure and based on that it will start the assginment.
Surely this region will not be in AM#this.servers as RS B is not the owner of
this.
To answer the 2nd question about OPENED regions. Again this should not be a
problem. There are 2 cases in this
1> OpenedRegionHandler is completed. -> Here this means that this.servers is
already updated. And the region plan is also cleared. Now when SSH starts it
will see the region in AM.this.servers and also there is nothing in RIT. So
SSH will scan the meta find this region and go and assign it to a new RS.
(Patch is not needed here)
2> OpenedRegionHandler is not yet completed. -> Here there may be case
OpenedRegionHandler has not yet done his work. So the regionPlan is still
available but the this.servers and this.regions are not yet updated. Now as
per patch we know that
{code}
if (rit != null && !rit.isClosing() && !rit.isPendingClose() &&
!rit.isSplitting()
+ && !regionsFromRegionPlansForServer.contains(rit.getRegion()))
{
{code}
The new datastructure contatins that region and hence we don not skip the
assignment and go ahead with assignment. This region got added here because we
have it in Regionplan with the destination as the dead server.
bq.we can do to better encapsulate this new state management?
Yes we will try to do that.
bq.Can't we just keep list of dead servers and any attempt at getting a region
plan related to the dead server or server being processed as dead is put off
because SSH will get to it?
Already we are doing this. But we cannot guarentee at what point the RS had
gone down, is it after forming the regionPlan or before forming the regionplan.
IF it is before forming the regionplan
we are already excluding the dead servers.
Pls correct me if am wrong and if you have any doubts
> Regions's in OPENING state from failed regionservers takes a long time to
> recover
> ---------------------------------------------------------------------------------
>
> Key: HBASE-6060
> URL: https://issues.apache.org/jira/browse/HBASE-6060
> Project: HBase
> Issue Type: Bug
> Components: master, regionserver
> Reporter: Enis Soztutar
> Assignee: rajeshbabu
> Fix For: 0.96.0, 0.94.1, 0.92.3
>
> Attachments: 6060-94-v3.patch, 6060-94-v4.patch, 6060-94-v4_1.patch,
> 6060-94-v4_1.patch, 6060-trunk.patch, 6060-trunk.patch, 6060-trunk_2.patch,
> 6060-trunk_3.patch, 6060_suggestion_based_off_v3.patch, HBASE-6060-92.patch,
> HBASE-6060-94.patch
>
>
> we have seen a pattern in tests, that the regions are stuck in OPENING state
> for a very long time when the region server who is opening the region fails.
> My understanding of the process:
>
> - master calls rs to open the region. If rs is offline, a new plan is
> generated (a new rs is chosen). RegionState is set to PENDING_OPEN (only in
> master memory, zk still shows OFFLINE). See HRegionServer.openRegion(),
> HMaster.assign()
> - RegionServer, starts opening a region, changes the state in znode. But
> that znode is not ephemeral. (see ZkAssign)
> - Rs transitions zk node from OFFLINE to OPENING. See
> OpenRegionHandler.process()
> - rs then opens the region, and changes znode from OPENING to OPENED
> - when rs is killed between OPENING and OPENED states, then zk shows OPENING
> state, and the master just waits for rs to change the region state, but since
> rs is down, that wont happen.
> - There is a AssignmentManager.TimeoutMonitor, which does exactly guard
> against these kind of conditions. It periodically checks (every 10 sec by
> default) the regions in transition to see whether they timedout
> (hbase.master.assignment.timeoutmonitor.timeout). Default timeout is 30 min,
> which explains what you and I are seeing.
> - ServerShutdownHandler in Master does not reassign regions in OPENING
> state, although it handles other states.
> Lowering that threshold from the configuration is one option, but still I
> think we can do better.
> Will investigate more.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
