Custom metrics are not showing up when parallel_worker=0

[saeid93]

Following the root cause of #1130 it seems that the MLServer metrics not working with the parallel_worker=0 was the reason. As a result Custom metrics seems to be not working in the Seldon core V1 wrapped MLServers when the parallel_worker>0.

[adriangonz]

Thanks for taking the time to troubleshoot this one!

There may be something strange going on with the base metrics register. We'll have a look :+1:

[Kolajik]

Hi @saeid93, does it work for you if you set parallel_workers to 0, though?

We're running barebone MLServer without Seldon Core inside of our own kubernetes clusters and either with MLSERVER_PARALLEL_WORKERS env variable set to 0 or >0, it does not seem to work.

However, it does register those metrics locally if we just start mlserver with mlserver start . command in CLI.

[Kolajik]

Hi @saeid93, does it work for you if you set parallel_workers to 0, though?

We're running barebone MLServer without Seldon Core inside of our own kubernetes clusters and either with MLSERVER_PARALLEL_WORKERS env variable set to 0 or >0, it does not seem to work.

However, it does register those metrics locally if we just start mlserver with mlserver start . command in CLI.

Our bad as we were running an old version of mlserver - 1.2.4 (custom metrics are available since 1.3.0). But still curious, does it work for you just fine? 🙂