Hello Team,
I observed this is Pega 8.4.2 and now in Pega 8.6.3.
The search indices on the filesystem are illegible and if we have just one kind of indices corrupted, then we are not able to isolate them and have to remove all the indices and rebuilt everything.
I recall there was a hotfix for this issue early on in Infinity releases, is there a known fix to this issue?
Best regards,
Pawan
Hello Team,
Anything on this?
Thanks in advance!
-Pawan
Hi @pawann I have gone through past support tickets and am unable to find any reference to hotfixes existing to remedy this particular symptom, even in 8.4.
I found a support article which shows workarounds when Work indexes display as unavailable on search landing page.
For earlier releases this was also documented: Batch creation and updating of Elasticsearch index files from the command line
For Pega 8.6 you can follow: Rebuilding search indexes by using a direct connection to the database and Checking search index status
All resolved issues for the latest 8.6x patch can be found here: Pega Platform 8.6.6 Patch Resolved Issues
I found one client who had a Cloud environment where the issue occurred because of a faulty configuration base where a cleanup of older nodes did not happen and it picked random util nodes to attach and caused ES corruption. Web node accidently updates replica count which creates impacts Search Stability.
That issue was resolved in 8.6.6 under BUG-747182
Replica count updated only for index nodes
Every time a search node is restarted or reinitialized, a refresh command (refreshClients command) is sent to all nodes. A web node which is not a search node will start a reinitialization designed to reattach the node to ES cluster, and it was expected that the web node would not change information about the replica count since it is not responsible for storing data. However, the non-search node was changing the replica count due to a condition created when locking was not ready on the search node. This has been resolved with an update to insure replica count is updated only for index nodes in PegaSearchProviderImpl::reInitializeESNode.
Therefore it might be worthwhile to update to 8.6.6 to test your use case.
Thanks @MarijeSchillern, at this moment we have no plans to upgrade to 866(not that we could do it just like that at the bank).
We however, will be moving to Pega 8.8, can you confirm if this is fixed in Pega 8.8?
Best regards,
Pawan
Hi . @pawann
We understand that major upgrades require a lot of planning and time. Patch updates, however, are considered an integral part of keeping Pega up-to-date. Pega Platform patches bundle the bug and security fixes over a given period into a cumulative release that improves the overall quality of the software without touching the environment infrastructure.
See Applying a patch without downtime
On-premises clients can apply patches by using Pega best practice to achieve zero downtime on your maintenance schedule. For more information on applying patches, see Apply the latest Pega Platform software patch. You can obtain patches from the My Pega Software portal.
See Apply the latest Pega Platform patch
If your issue is due to the replicate count issue I described, then I can confirm that it was fixed in 8.8.1 under BUG-757787.
