RAM consumption issues

[apneadiving]

Hi and thanks for the gem :)

I wrote before to the support but they made me bounce back here.

We have just switched to the google ads gem vs the adwords gem we used before.
Our codebase has not changed except the specific library implementation.
We noticed faster sync times which is good.
However, we now have a surge in RAM usage. We used to have flat 500MB processes but now our RAM usage is over the top and we had to change our infrastructure to be able to deal with this.
Below are screenshots of our RAM monitoring process per process and globally. You can easily guess from it when we deployed the google ads implementation.
It seems protobuf coud be a reason

Are you aware of such a behavior in the gem? If so, do you have plans/deadlines to work on it?

Thanks

[apneadiving]

We switched to the 3.15.8 version of google-protobuf this morning.
It looks promising

[mcloonan]

Glad to hear that the newer version of protobuf is helping. I'll get in touch with the protobuf team here to see if this was expected and what the long-term plans are.

I'll see if I can run some analysis to see how much memory is being used by various pieces, so we can determine whether there are any improvements we can make on our layer as well.

[haberman]

There was a significant memory leak in Protobuf Ruby 3.15.0 through 3.15.7 that was fixed in 3.15 8: protocolbuffers/protobuf#8421.

[apneadiving]

I confirm it is flat under 500MB since we changed protobuf version.

[apneadiving]

[apneadiving]

kinda unrelated but along with the (now solved) RAM issue, we have now some workers hanging indefinitely. Is there some retry mechanism for instance which could explain this?
Thanks :)

[mcloonan]

Probably the same as issue #246