Thanks Thejas for this patch! I'm also going to restart PTest and force recreation of the test slaves based on fresh image to see if it resolves the issue (in case the test slaves are getting too loaded and slow to start HMS in time). If not then Thejas's patch should tell us a bit more.
On Tue, Apr 12, 2016 at 12:35 AM, Thejas Nair <thejas.n...@gmail.com> wrote: > Created a patch that should hopefully help in figuring out whats going > on - https://issues.apache.org/jira/browse/HIVE-13491 > > > On Wed, Apr 6, 2016 at 1:56 PM, Szehon Ho <sze...@cloudera.com> wrote: > > Yea thanks for point it out. I see it too and am not able to reproduce > it > > locally. It points to an environment issue, but not aware anything > changed > > with the environment. > > > > Anyone have any ideas? > > > > On Wed, Apr 6, 2016 at 1:29 PM, Sergey Shelukhin <ser...@hortonworks.com > > > > wrote: > > > >> Has anyone else noticed that many tests that involve metastore started > >> failing lately? The failures are sporadic and happen both in the tests > >> that test metastore, and q files that use metastore… > >> The error is always something like > >> java.net.ConnectException: Connection refused > >> at java.net.PlainSocketImpl.socketConnect(Native Method) > >> at > >> > java.net.AbstractPlainSocketImpl.doConnect(AbstractPlainSocketImpl.java:339 > >> ) > >> at > >> > java.net.AbstractPlainSocketImpl.connectToAddress(AbstractPlainSocketImpl.j > >> ava:198) > >> at > >> > java.net.AbstractPlainSocketImpl.connect(AbstractPlainSocketImpl.java:182) > >> at java.net.SocksSocketImpl.connect(SocksSocketImpl.java:392) > >> at java.net.Socket.connect(Socket.java:579) > >> at > >> > org.apache.hadoop.hive.metastore.MetaStoreUtils.loopUntilHMSReady(MetaStore > >> Utils.java:1208) > >> at > >> > org.apache.hadoop.hive.metastore.MetaStoreUtils.startMetaStore(MetaStoreUti > >> ls.java:1195) > >> at > >> > org.apache.hadoop.hive.metastore.TestMetaStoreMetrics.before(TestMetaStoreM > >> etrics.java:54) > >> > >> I wonder if someone has insight on whether this is an environment issue, > >> or someone broke something recently, before we investigate more :) > >> > >> > >> >
