On 5/30/23 18:48, Shawn Heisey wrote:
I am trying one more thing ... setting the client to use http/1.1. I
remember we had some issues with Solr where http2 wasn't working right,
and I think the final fix for that was upgrading Jetty.
This did not help. It still randomly (though not very frequently)
throws that exception about authentication.
I do think I was on the right track, just not looking in exactly the
right place.
I set an option when building the Solr client for indexing that SHOULD
have forced all updates to be routed to shard leaders.
Turns out that option doesn't actually do anything. Which I think is a bug.
So sometimes updates went to a replica that wasn't leader. And when
that happens, Solr forwards the request to the right index. That
forwarding is done using HTTP/2, and we have seen evidence that HTTP/2
on Solr sometimes has weird issues.
There is a different and very similarly named option that DOES force the
client to send to leaders only.
So I am using http/1.1 in conjunction with that correct option. Bonus
... the indexing is happening faster than before, because Solr doesn't
ever need to forward requests to another server. I should find out in
the next hour or so whether it actually helped.
I do think that if I can get them to upgrade Solr so it's using Jetty
10, the problem will evaporate.
Thanks,
Shawn
_______________________________________________
jetty-users mailing list
jetty-users@eclipse.org
To unsubscribe from this list, visit
https://www.eclipse.org/mailman/listinfo/jetty-users
