On 11/29/2018 5:27 AM, Ana Mercan (RO) wrote:
I have tried to subscribe to general forums but i wasn't able to, hence i would appreciate if we can keep this thread and have the replies over email.
The project does not have forums. Only mailing lists and an IRC channel. There are third parties that mirror this mailing list in a forum format, and a lot of people access the mailing list that way, but in general we recommend a direct subscription to the list. If you choose to access the list via a third-party website, you will need to talk to the people who run that website for help.
http://lucene.apache.org/solr/community.html#mailing-lists-irc
I want to have more background about the dependencies of solr/zookeeper services and the task schedulers -ful/incremental indexing.
Solr does not have any scheduling capabilities. If you have dataimport scheduled, that is going to be somewhere other than Solr. There is no intent to add scheduling capabilities to Solr, because ALL modern operating systems have job scheduling capability built in.
So in case of a manual restart or unplanned reboot our services are set to start manually hence i wanted to setup the dependency between them so I wanted to confirm if the order is the following one: 1.First Zookeeper service 2.Solr service
If you have a fully redundant zookeeper ensemble, it will consist of at least three separate servers, each running on separate physical hardware. If a majority of ZK servers is running with that server down, it won't matter whether you start ZK or Solr first ... but you probably should start ZK first, just in case you *don't* have a majority running.
If there is any documentation on this matter please let me know.
https://lucene.apache.org/solr/guide/7_5/setting-up-an-external-zookeeper-ensemble.html#how-many-zookeeper-nodes The reference guide is a big document, available in PDF.
Another similar topic is the following one: Yesterday we encountered a dirty reboot and right after the incremental indexing task was failing with : " Job Id: 172605, Status: Failed Indexing failed. Refer incremental-indexing.log for more information.
That error message did not come from Solr, unless you have modified your Solr extensively. Solr does not create a logfile named "incremental-indexing.log".
java.lang.Exception: File ./index.properties does not exist to read the property "last_index_time""
We would need to see the *ENTIRE* error. Java exceptions can be dozens of lines long.
The file was there but was empty without any timestamp So i did the following: Do the following: 1.open it check if is empty or not and take a backup of it 2.disable the task, take the content of that file from stage 3.check in the database the last successful one --timestamp 4.updat that file with the above timestamp 5.probably it wont work 6.if doesn t work update the timestamp with the current date and run the job,, once got successfully completed end it disable it and update the timestamp with the value from 2nd step 5 update [X].[dbo].[C_CACHE_SEARCH_INDEX_JOB_AUDIT] set Status = 'Failed' where Status ='In Progress'
Solr has no tasks. That looks like something related to Microsoft's SQL server, not Solr.
6.enable the task and run it and also monitor it will complete successfully
Again, Solr has no tasks. Whatever you're talking about here must be something other than Solr.
So my question is when this scenario can happen and also is there any procedure to stop/start the components in a controlled way?
Solr is stopped/started as a single unit -- there's no way to only do part of it or specify the component start order.
Thanks, Shawn
