[
https://issues.apache.org/jira/browse/HBASE-12844?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Enis Soztutar updated HBASE-12844:
----------------------------------
Attachment: hbase-12844_v2-0.98.patch
hbase-12844_v2-branch-1.patch
Attaching final patches.
bq. You've looked at what would be involved having only SM dead servers list
and having RS ask it for deadservers and too much work for 1.0?
Forgot to reply to this. I've poked around, but this scared me out of pursuing
that approach for now:
{code}
* A region server could be dead but we don't know it yet. We may
* think it's online falsely. Therefore if a server is online, we still
* need to confirm it reachable and having the expected start code.
*/
synchronized boolean wasRegionOnDeadServer(final String encodedName) {
{code}
> ServerManager.isServerReacable() should sleep between retries
> -------------------------------------------------------------
>
> Key: HBASE-12844
> URL: https://issues.apache.org/jira/browse/HBASE-12844
> Project: HBase
> Issue Type: Bug
> Reporter: Enis Soztutar
> Assignee: Enis Soztutar
> Fix For: 1.0.0, 2.0.0, 0.98.10, 1.1.0
>
> Attachments: HBASE-12844-0.98.patch, HBASE-12844-0.98.patch,
> hbase-12844_v1.patch, hbase-12844_v2-0.98.patch,
> hbase-12844_v2-branch-1.patch, hbase-12844_v2.patch
>
>
> There is a fundamental problem with the way assignment manager and cluster
> membership works. Basically, the root cause of most of the complexity and
> root cause for many bugs is that we do have multiple "cluster membership"
> sources. This causes problems when they diverge from each other.
> Master's in-memory ServerManager class keep track of what servers are online
> and what servers are considered dead. We have online and dead servers list in
> ServerManager, and a separate dead servers list in RegionStates.
> There are at least 3 ways that a server can join into the dead list. First is
> the zookeeper session. If a server loses it's zk session, the master gets
> notification and expires the server. This is the regular way.
> Second is calls through ServerManager.expireServer(). On master this is
> mostly through master rejoining the cluster. Master waits for some time for
> RS's to heartbeat and expires all others and process them as dead servers.
> This method has the potential to hijack the regions in a region server
> without the region server knowing about it (and thus can cause multi homing
> of regions for reads etc).
> Third is the RegionStates calling ServerManager.isServerReachable() and if
> not adding the server to it's own dead list, but not to the dead list of
> ServerManager.
> Obviously, as in the region assignment case as well as this, we should fix
> the "state is kept in multiple places" syndrome, but not in this issue (we
> already have HBASE-5487, etc for that).
> In this issue we should at least solve the following case:
> When a region server is starting up, it will throw exceptions when we want to
> ping:
> {code}
> 2015-01-10 00:23:10,369 DEBUG [AM.-pool1-t5] master.ServerManager: Couldn't
> reach os-enis-hbase-1.0-test-1.hw.com,16020,1420849386091, try=0 of 10
> org.apache.hadoop.hbase.ipc.ServerNotRunningYetException:
> org.apache.hadoop.hbase.ipc.ServerNotRunningYetException: Server is not
> running yet
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.checkOpen(RSRpcServices.java:886)
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.getServerInfo(RSRpcServices.java:1155)
> at
> org.apache.hadoop.hbase.protobuf.generated.AdminProtos$AdminService$2.callBlockingMethod(AdminProtos.java:20886)
> at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2028)
> at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:108)
> at
> org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:130)
> at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:107)
> at java.lang.Thread.run(Thread.java:745)
> at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native
> Method)
> at
> sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:57)
> at
> sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
> at java.lang.reflect.Constructor.newInstance(Constructor.java:526)
> at
> org.apache.hadoop.ipc.RemoteException.instantiateException(RemoteException.java:106)
> at
> org.apache.hadoop.ipc.RemoteException.unwrapRemoteException(RemoteException.java:95)
> at
> org.apache.hadoop.hbase.protobuf.ProtobufUtil.getRemoteException(ProtobufUtil.java:309)
> at
> org.apache.hadoop.hbase.protobuf.ProtobufUtil.getServerInfo(ProtobufUtil.java:1794)
> at
> org.apache.hadoop.hbase.master.ServerManager.isServerReachable(ServerManager.java:810)
> at
> org.apache.hadoop.hbase.master.RegionStates.isServerDeadAndNotProcessed(RegionStates.java:756)
> at
> org.apache.hadoop.hbase.master.AssignmentManager.forceRegionStateToOffline(AssignmentManager.java:1952)
> at
> org.apache.hadoop.hbase.master.AssignmentManager.assign(AssignmentManager.java:1559)
> at
> org.apache.hadoop.hbase.master.AssignCallable.call(AssignCallable.java:48)
> at java.util.concurrent.FutureTask.run(FutureTask.java:262)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> at java.lang.Thread.run(Thread.java:745)
> Caused by:
> org.apache.hadoop.hbase.ipc.RemoteWithExtrasException(org.apache.hadoop.hbase.ipc.ServerNotRunningYetException):
> org.apache.hadoop.hbase.ipc.ServerNotRunningYetException: Server is not
> running yet
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.checkOpen(RSRpcServices.java:886)
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.getServerInfo(RSRpcServices.java:1155)
> at
> org.apache.hadoop.hbase.protobuf.generated.AdminProtos$AdminService$2.callBlockingMethod(AdminProtos.java:20886)
> at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2028)
> at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:108)
> at
> org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:130)
> at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:107)
> at java.lang.Thread.run(Thread.java:745)
> at
> org.apache.hadoop.hbase.ipc.RpcClientImpl.call(RpcClientImpl.java:1199)
> at
> org.apache.hadoop.hbase.ipc.AbstractRpcClient.callBlockingMethod(AbstractRpcClient.java:216)
> at
> org.apache.hadoop.hbase.ipc.AbstractRpcClient$BlockingRpcChannelImplementation.callBlockingMethod(AbstractRpcClient.java:300)
> at
> org.apache.hadoop.hbase.protobuf.generated.AdminProtos$AdminService$BlockingStub.getServerInfo(AdminProtos.java:21819)
> at
> org.apache.hadoop.hbase.protobuf.ProtobufUtil.getServerInfo(ProtobufUtil.java:1791)
> ... 9 more
> ....
> 2015-01-10 00:23:10,399 DEBUG [AM.-pool1-t5] master.ServerManager: Couldn't
> reach os-enis-hbase-1.0-test-1.hw.com,16020,1420849386091, try=9 of 10
> org.apache.hadoop.hbase.ipc.ServerNotRunningYetException:
> org.apache.hadoop.hbase.ipc.ServerNotRunningYetException: Server is not
> running yet
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.checkOpen(RSRpcServices.java:886)
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.getServerInfo(RSRpcServices.java:1155)
> at
> org.apache.hadoop.hbase.protobuf.generated.AdminProtos$AdminService$2.callBlockingMethod(AdminProtos.java:20886)
> at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2028)
> at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:108)
> at
> org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:130)
> at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:107)
> at java.lang.Thread.run(Thread.java:745)
> at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native
> Method)
> at
> sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:57)
> at
> sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
> at java.lang.reflect.Constructor.newInstance(Constructor.java:526)
> at
> org.apache.hadoop.ipc.RemoteException.instantiateException(RemoteException.java:106)
> at
> org.apache.hadoop.ipc.RemoteException.unwrapRemoteException(RemoteException.java:95)
> at
> org.apache.hadoop.hbase.protobuf.ProtobufUtil.getRemoteException(ProtobufUtil.java:309)
> at
> org.apache.hadoop.hbase.protobuf.ProtobufUtil.getServerInfo(ProtobufUtil.java:1794)
> at
> org.apache.hadoop.hbase.master.ServerManager.isServerReachable(ServerManager.java:810)
> at
> org.apache.hadoop.hbase.master.RegionStates.isServerDeadAndNotProcessed(RegionStates.java:756)
> at
> org.apache.hadoop.hbase.master.AssignmentManager.forceRegionStateToOffline(AssignmentManager.java:1952)
> at
> org.apache.hadoop.hbase.master.AssignmentManager.assign(AssignmentManager.java:1559)
> at
> org.apache.hadoop.hbase.master.AssignCallable.call(AssignCallable.java:48)
> at java.util.concurrent.FutureTask.run(FutureTask.java:262)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> at java.lang.Thread.run(Thread.java:745)
> Caused by:
> org.apache.hadoop.hbase.ipc.RemoteWithExtrasException(org.apache.hadoop.hbase.ipc.ServerNotRunningYetException):
> org.apache.hadoop.hbase.ipc.ServerNotRunningYetException: Server is not
> running yet
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.checkOpen(RSRpcServices.java:886)
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.getServerInfo(RSRpcServices.java:1155)
> at
> org.apache.hadoop.hbase.protobuf.generated.AdminProtos$AdminService$2.callBlockingMethod(AdminProtos.java:20886)
> at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2028)
> at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:108)
> at
> org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:130)
> at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:107)
> at java.lang.Thread.run(Thread.java:745)
> at
> org.apache.hadoop.hbase.ipc.RpcClientImpl.call(RpcClientImpl.java:1199)
> at
> org.apache.hadoop.hbase.ipc.AbstractRpcClient.callBlockingMethod(AbstractRpcClient.java:216)
> at
> org.apache.hadoop.hbase.ipc.AbstractRpcClient$BlockingRpcChannelImplementation.callBlockingMethod(AbstractRpcClient.java:300)
> at
> org.apache.hadoop.hbase.protobuf.generated.AdminProtos$AdminService$BlockingStub.getServerInfo(AdminProtos.java:21819)
> at
> org.apache.hadoop.hbase.protobuf.ProtobufUtil.getServerInfo(ProtobufUtil.java:1791)
> ... 9 more
> {code}
> After 10 attempts happening in 10s of milliseconds (as opposed to sleeping
> between retries, the server is put in the dead servers list in RegionStates
> (but not in ServerManager's dead servers list). This results in the region
> server never receiving YouAreDeadException, and the ServerManager thinking
> that the server is alive and well, while the RegionStates thinks that the RS
> is dead and not assigning regions:
> {code}
> 2015-01-10 00:23:13,163 INFO
> [os-enis-hbase-1.0-test-2.hw.com,16020,1420848162613-GeneralBulkAssigner-0]
> master.AssignmentManager: Assigning 2 region(s) to
> os-enis-hbase-1.0-test-1.hw.com,16020,1420849386091
> 2015-01-10 00:23:13,170 WARN
> [os-enis-hbase-1.0-test-2.hw.com,16020,1420848162613-GeneralBulkAssigner-0]
> master.RegionStates: Couldn't reach online server
> os-enis-hbase-1.0-test-1.hw.com,16020,1420849386091
> {code}
> This also prevents unassign etc leaving the regions in transition state
> forever (until the admin kills the RS manually).
> {code}
> 2015-01-10 00:23:13,188 INFO
> [os-enis-hbase-1.0-test-2.hw.com,16020,1420848162613-GeneralBulkAssigner-0]
> master.AssignmentManager: Skip assigning
> loadtest_d1,cccccccc,1420849388510.15a752a6ad4b3a21c0d471483a225144., it is
> on a dead but not processed yet server:
> os-enis-hbase-1.0-test-1.hw.com,16020,1420849386091
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
