Solr6 Reindex - unhealthy Container

[maxodoble]

Hi,

when testing an update to Alfresco Community 25.1.0 i am noticing the AIO Transform Container going to an unhealthy status real quick (a few minutes), when performing a complete Solr FTS Reindex.

calling the health endpoint manually shows something like: "maxTransform limit reached (10240), container wants to be killed."

is `maxTransforms` configurable via env var in the compose service for the aio transformer? especially during a reindex of a million documents, i'd like to keep  a healthy status of the transform container, and not going to unhealthy after transforming about 10.000 docs.

or can the unhealthy status be ignored?

[LeoMattioli · Accepted solution]

You can set it to 0 to disable the check

Here's the docs with more explanation: https://github.com/Alfresco/alfresco-transform-core/blob/master/docs/Probes.md

Hope this helps.

[maxodoble]

Thanks, will try that out!

I'd suggest though, to eventually make the documentation you linked to, a little bit more precise or clear:

I initially thought, this doc doesn't apply to deployment with docker compose because of the title: "Transformer k8s liveness and readiness probes", that i interpretated as "ah... K8s only".

The documentation seems to be out of date too: "The T-Engine liveness probes will be reevaluated/changed/improved as part of the ATS-138 story.", because the probes seem to be enabled by default now, otherwise i wouldn't have stumbeled upon this topic with my tests.

Please let me know, if this is documented anywhere else already, in case i have missed this.

Thanks again,

Max

[maxodoble]

Hello again,

after trying out a complete reindex with this env-var: maxTransform = 0:

Transform container stayed "healthy" for much longer, but after about half an hour the transform Container was restarted by docker (for whatever reason i couldn't see). and after the container was restarted, the unhealthy status came up rather quickly again, and even the maxTransforms exceeded (10.240) message showed up again in the logs.

the env-var was still active (checked inside the transform) container.

manually checking the health endpoint for the transform container with curl: HTTP 409.

two questions:

* would it help to open a github issue in transform aio project?

* i suspect that because of the container restarts during indexing some docs are not fts indexed. would initiating a index fix possibly reindex these, so that all docs are in the index?

[maxodoble]

sorry, typo above, not 409, but 429:

manually checking the health endpoint for the transform container with curl: HTTP 429 (too many requests)

[LeoMattioli]

Hi,

* would it help to open a github issue in transform aio project? Probably yes, however I don't know why the maxTransform configuration goes back to 10.240 (but I'm not a devOps)

* i suspect that because of the container restarts during indexing some docs are not fts indexed. would initiating a index fix possibly reindex these, so that all docs are in the index? Yes, you can use action reindex single nodeIds or even using a query (see docs)

Just adding one little note: be aware that Solr in Search Services is not deployable in k8s as the indexes stored by the Solr are locked by the application and cannot be reused for other Solr copies of the app.