And are you saying you already built the latest -dev and then the customer was mailing?
no Will do that the new build tonight
Really as much of How to Get Help - KumoMTA Docs as you can gather is ideal.
will do .. that
Also going to need an idea of the queue depth around the time of the issue. You may want to continuously monitor the metrics API to see how large things get before the lockup.
I’ve seen this happen 2 or 3 times as well, and I also suspected it was related to log webhooks, but I hasn’t happened for me again in over a month.
well apart from that I am running it on a 4Gb RAM server… I might just get one with 8Gb to see if that makes a difference.
there are memory related metrics returned in the prometheus metrics that you should probably keep an eye on if you suspect that you are using up too much ram; memory_over_limit_count, memory_usage and memory_limit.
The limit is configured at the system level, either via cgroups or ulimit.
systemctl status kumomta I have seen on 1.1Gb but most of the time it is just above 500Mb
but I will check the metrics and see if I can monitor those
Hmmm I upgraded the server to 8Gb on AWS… just to count out memory issues with Logstash and kumomta on this Debina 12 server. But after the latest version build on a Debian 12 host was started. it was all fine
but when I let some mails from active mailing customers flow in it crashed right away.
i do notice errors on trying to deliver on ipv6 addr5esses to Google
what precisely do you mean by “crashed right away”?
I was tracing smtp… and saw a new mail being accepted and then the kli strace is disconnected
[172.31.45.163:60540->172.31.17.90:2625] 54ms <- 354 Send body; end with CRLF.CRLF
[172.31.45.163:60548->172.31.17.90:2625] 0ns === Connected 2024-03-20 21:11:29.405658154 UTC
[172.31.45.163:60548->172.31.17.90:2625] 0ns === conn_meta hostname="smtp-out3.mxtochina.com"
[172.31.45.163:60548->172.31.17.90:2625] 0ns === conn_meta received_from="172.31.45.163:60548"
[172.31.45.163:60548->172.31.17.90:2625] 0ns === conn_meta received_via="172.31.17.90:2625"
[172.31.45.163:60548->172.31.17.90:2625] 0ns === conn_meta reception_protocol="ESMTP"
[172.31.45.163:60548->172.31.17.90:2625] 54µs <- 220 smtp-out3.mxtochina.com KumoMTA
Error: WebSocket protocol error: Connection reset without closing handshake
Caused by:
Connection reset without closing handshake
but in this case I don’t get the dump of the mail content..
and the crash is actually a hang
so nothing responds and I guess that the systemctl stop kumomta kills the process in the end
I just build the lastest version: ii kumomta-dev 2024.03.20.082746.c44d7fee amd64 A high performance, modern MTA
Hmm… is that a kcli error or a KumoMTA error?