On 10/25/2017 6:44 PM, Tarjono, C. A. wrote: > Thanks so much for your input! We will try your suggestion and hope it will > resolve the issue. > > On the side note, would you know if this is an existing bug? if yes, has it > been resolved in later version? i.e. zk allows adding nodes when it exceeds > the buffer. > > We are currently using ZK 3.4.6 to use with SolrCloud 5.1.0.
The ZOOKEEPER-1162 issue has not been fixed. It is a very old bug -- opened six years ago. They probably aren't going to fix it. If you find that restarting a single Solr instance ends up filling the queue with too many entries, you may need to increase the jute.maxbuffer setting on both Solr and ZK so that a large queue won't cause everything to break. There has been some effort in recent 6.x versions to improve this situation, as Erick mentioned in his reply. There's nothing that can be done for problems like this in 5.x versions. Thanks, Shawn
