This message was deleted.
# ask-for-helps
Slackbot
10/16/2023, 12:56 PMThis message was deleted.
m
Michael Aydinbas
10/16/2023, 1:21 PMAnd just to make clear how urgend and problematic that is: the container started initially with 3.5 GB used memory...so we are already +1.5 GB
j
Jian Shen Yap
10/16/2023, 9:53 PMHi @Michael Aydinbas, thanks for finding this information.
Do you see the memory used increases over time with requests?
Is it possible to simulate this behaviours by firing alot requests to the deployment?
m
Michael Aydinbas
10/17/2023, 1:16 PMwhat exactly do you mean? Yes, the service is receiving request so its not idle behavior that is true but the number of requests is really low, peaks with 20 requests/second max
Michael Aydinbas
10/17/2023, 1:17 PMAnd I am also really worried about the latency of the app, like the 99 percentile is showing request durations always in the range of 2-3 seconds, which is much to high for a single SVM, so on my computer locally this app has a latency of 200 ms but I guess it has somehow to do with workers being too busy
Michael Aydinbas
10/18/2023, 5:56 PMCould I get some help here, pls?
j
Jian Shen Yap
10/18/2023, 11:09 PMHey @Michael Aydinbas what I meant was, if you could help us to generate a simple reproducible example to simulate the memory build up issue, that would be helpful for us to track it!
as for the latency, did you enabled adaptive batching? You could set the
max_latency