On 10/22/2018 3:31 PM, Susheel Kumar wrote:
Hello,
I am seeing "ZookeeperServer not running" WARM messages in zookeeper logs
which is causing the Solr client connections to timeout...
What could be the problem?
ZK: 3.4.10
Zookeeper.out
==
For help with the ZK server log, you'll need to consult the ZooKeeper
project. The language in their log entries seems plain enough, but they
will be able to tell you precisely what it means.
solr.log
2018-10-22 10:02:21.466 WARN (main-SendThread(srch0118:2182)) [ ]
o.a.z.ClientCnxn Client session timed out, have not heard from server in
26675ms for sessionid 0x5665c67cb0d000b
The ZK client in Solr hasn't heard from the ZK server in over 26
seconds, so it considers that connection to have timed out, and will
throw the connection away. It should try again to establish a new
connection ... but whatever's causing the problem will probably also
affect the new connection.
It's a particularly bad sign for the ZK connection to time out,
especially on an interval like 26 seconds. That's a REALLY long time
for software like Solr and ZK.
One of the things that can cause problems like this is having a heap
that's too small, so Java must spend the majority of its time doing
garbage collection, rather than running the program it's been asked to
run. There are sometimes other causes, but that is a very common cause.
Can you share a garbage collection log from a time when these errors
happen? Solr should set up Java so that it creates a GC log. You'll
need to use a file sharing site (like Dropbox) -- email attachments
almost never make it to the list.
Thanks,
Shawn
