Hi Teams
Today, I found that kumo’s memory is full of memory. How can I start to find out what causes the excessive memory occupation.
Due to the data quality problem sent by one of our customers, there are many in particular:
KumoMTA internal: no sources for xxxx pool=‘xxxx’ are eligible for selection at this time
Could this have something to do with it?
There should be memory stats in that list as well as lua context use.
You may want to look at the metrics feed.
https://docs.kumomta.com/reference/http/metrics/
If you use prometheus and grafana, you can get some great visuals on resource usage
These may help as well:
Thanks Tom. Ihave use metrics and prometheus , I can also get how much memory is occupied, but I have a total of four kumo, three of which have large memory fluctuations. I want to know the reason, because the memory quickly reaches 100 percent, and I am not sure whether it will affect sending mail.
Your top also indicates you have 409k resident messages. Assuming your average message sise is 100k, then that alone will account for 40Gb (? is that math right?)
Yes, right, what kind of message is resident messages?
There is indeed an IP ban caused by Spam in the customer’s data, and we are communicating with the customer.
Does he store all the mail caused by Spam into memory? My original understanding was to store it on disk.
You may want to check the max_ready value for that user
Can I limit the max_ready of this tenant?
A resident message is one that is loaded into memory waiting to deliver - typically these are in the ready queue. If you are not able to deliver, but also not getting a failure message then these will just build up.
That depends on how you have them configured.
it is set as shaping
But that is only one possibility,
You might have complex Lua that is using and not releaseing resources
I use helper , I will test it in shaping.toml to limit it .
for the doc ,it said : The default for max_ready is 1024 messages. .. far more on my server