We did some more digging into this and here is the theory.
1. The regionservers use their local ip to lookup their hostnames and pass that
to the HMaster. The HMaster finds the server info by using this hostname as the
key in the HashMap.
HRegionServer.java
reinitialize() -
this.serverInfo = new HServerInfo(new HServerAddress(
new InetSocketAddress(address.getBindAddress(),
this.server.getListenerAddress().getPort())), System.currentTimeMillis(),
this.conf.getInt("hbase.regionserver.info.port", 60030), machineName);
In run() -
HMsg msgs[] = hbaseMaster.regionServerReport(
serverInfo, outboundArray, getMostLoadedRegions());
2. I have observed in the past that there could be some DNS flakiness which
causes the IP address of the machines to be returned as their hostnames.
Guessing this is what happened.
3. The HMaster looks in the map for the above IP address (masquerading as the
server name). It gets and does not find the entry in its map. So it assumes
that this is a new region server and issues a CALL_SERVER_STARTUP.
4. The region server that receives it is in fact already running (under its
real hostname) and enters the "HMaster panic" mode and bad stuff happens.
ServerManager.java in regionServerReport() -
HServerInfo storedInfo = serversToServerInfo.get(info.getServerName());
if (storedInfo == null) {
// snip...
return new HMsg[] {CALL_SERVER_STARTUP};
}
Any reason why we use the hostname instead of the ip address in the map that
stores the regionserver info?
Thanks
Karthik
-----Original Message-----
From: Karthik Ranganathan [mailto:[email protected]]
Sent: Thursday, January 28, 2010 3:58 PM
To: [email protected]
Subject: Cannot locate root region
Hey guys,
Ran into some issues while testing and wanted to understand what has happened
better. Got the following exception when I went to the web UI
Trying to contact region server 10.129.68.204:60020 for region .META.,,1, row
'', but failed after 3 attempts.
Exceptions:
org.apache.hadoop.hbase.NotServingRegionException:
org.apache.hadoop.hbase.NotServingRegionException: .META.,,1
at
org.apache.hadoop.hbase.regionserver.HRegionServer.getRegion(HRegionServer.java:2254)
at
org.apache.hadoop.hbase.regionserver.HRegionServer.openScanner(HRegionServer.java:1837)
at sun.reflect.GeneratedMethodAccessor4.invoke(Unknown Source)
at
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
at java.lang.reflect.Method.invoke(Method.java:597)
at org.apache.hadoop.hbase.ipc.HBaseRPC$Server.call(HBaseRPC.java:648)
at
org.apache.hadoop.hbase.ipc.HBaseServer$Handler.run(HBaseServer.java:915)
>From a program that reads from a HBase table:
java.lang.reflect.UndeclaredThrowableException
at $Proxy1.getRegionInfo(Unknown Source)
at
org.apache.hadoop.hbase.client.HConnectionManager$TableServers.locateRootRegion(HConnectionManager.java:985)
at
org.apache.hadoop.hbase.client.HConnectionManager$TableServers.locateRegion(HConnectionManager.java:625)
at
org.apache.hadoop.hbase.client.HConnectionManager$TableServers.locateRegion(HConnectionManager.java:601)
at
org.apache.hadoop.hbase.client.HConnectionManager$TableServers.locateRegionInMeta(HConnectionManager.java:675)
<snip>
Followed up on the hmaster's log:
2010-01-28 11:21:16,148 INFO org.apache.hadoop.hbase.master.BaseScanner:
RegionManager.metaScanner scan of 1 row(s) of meta region {server:
10.129.68.204:60020, regionname: .META.,,1, startKey: <>} complete
2010-01-28 11:21:16,148 INFO org.apache.hadoop.hbase.master.BaseScanner: All 1
.META. region(s) scanned
2010-01-28 11:21:34,539 DEBUG org.apache.hadoop.hbase.master.ServerManager:
Received report from unknown server -- telling it to MSG_CALL_SERVER_STARTUP:
10.129.68.203,60020,1263605543210
2010-01-28 11:21:35,622 INFO org.apache.hadoop.hbase.master.ServerManager:
Received start message from: hbasetest004.ash1.facebook.com,60020,1264706494600
2010-01-28 11:21:36,649 DEBUG
org.apache.hadoop.hbase.zookeeper.ZooKeeperWrapper: Updated ZNode
/hbase/rs/1264706494600 with data 10.129.68.203:60020
2010-01-28 11:21:40,704 INFO org.apache.hadoop.ipc.HBaseServer: IPC Server
handler 39 on 60000, call createTable({NAME => 'test1', FAMILIES => [{NAME =>
'cf1', VERSIONS => '3', COMPRESSION => 'NONE', TTL => '2147483647', BLOCKSIZE
=> '65536', IN_MEMORY => 'false', BLOCKCACHE => 'true'}]}) from
10.131.29.183:63308: error: org.apache.hadoop.hbase.TableExistsException: test1
org.apache.hadoop.hbase.TableExistsException: test1
at org.apache.hadoop.hbase.master.HMaster.createTable(HMaster.java:792)
at org.apache.hadoop.hbase.master.HMaster.createTable(HMaster.java:756)
at sun.reflect.GeneratedMethodAccessor6.invoke(Unknown Source)
at
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
at java.lang.reflect.Method.invoke(Method.java:597)
at org.apache.hadoop.hbase.ipc.HBaseRPC$Server.call(HBaseRPC.java:648)
at
org.apache.hadoop.hbase.ipc.HBaseServer$Handler.run(HBaseServer.java:915)
>From a hregionserver's logs:
2010-01-28 11:20:22,589 DEBUG org.apache.hadoop.hbase.io.hfile.LruBlockCache:
Cache Stats: Sizes: Total=19.661453MB (20616528), Free=2377.0137MB
(2492479408), Max=2396.675MB (2513095936), Counts: Blocks=0, Access=0, Hit=0,
Miss=0, Evictions=0, Evicted=0, Ratios: Hit Ratio=NaN%, Miss Ratio=NaN%,
Evicted/Run=NaN
2010-01-28 11:21:22,588 DEBUG org.apache.hadoop.hbase.io.hfile.LruBlockCache:
Cache Stats: Sizes: Total=19.661453MB (20616528), Free=2377.0137MB
(2492479408), Max=2396.675MB (2513095936), Counts: Blocks=0, Access=0, Hit=0,
Miss=0, Evictions=0, Evicted=0, Ratios: Hit Ratio=NaN%, Miss Ratio=NaN%,
Evicted/Run=NaN
2010-01-28 11:22:18,794 INFO
org.apache.hadoop.hbase.regionserver.HRegionServer: MSG_CALL_SERVER_STARTUP
The code says the following:
case MSG_CALL_SERVER_STARTUP:
// We the MSG_CALL_SERVER_STARTUP on startup but we can also
// get it when the master is panicking because for instance
// the HDFS has been yanked out from under it. Be wary of
// this message.
Any ideas on what is going on? The best I can come up with is perhaps a flaky
DNS - would that explain this? This happened on three of our test clusters at
almost the same time. Also, what is the most graceful/simplest way to recover
from this?
Thanks
Karthik
