The company I work for are using a ChirpStack v3.8 for production.
Today, the app server was unresponsive. Particularly, when trying to access the Web GUI the browser timed out, and all API calls timed out as well.
We executed [journalctl -u chirpstack-application-server -f -n 50] and the last log was:
[Oct 10 09:23:57 ubuntu-8gb-nbg1-1 chirpstack-application-server: 2023/10/10 09:23:57 http: Accept error: accept tcp [::]:8003: accept4: too many open files; retrying in 1s]
Could you please offer some support for this? Restarting the service solved the problem but a crash occurred last week as well, which indicates that the issue is ongoing.
ChirpStack Application Server v3.8 was released on Feb 6, 2020, the ChirpStack Network Server v3.8 was also released in that same month. Since then many improvements and fixes have been implemented. I remember there was an issue with file-descriptors in the past, but I don’t know which in which version this was fixed. You will probably find this in the changelog.
In general, I would recommend looking into upgrading your ChirpStack installation, ideally (eventually) to v4.
ChirpStack v4 (which combines the NS + AS) was released about a year ago which also bring many improvements and simplifies the architecture
On top of that, I think hard upgrading an infrastructure as important to the LoRa environment as the Network Server is not the solution, fixing the issue is.
In any case v3.8 is very stable and I would like to fix it. I will be upgrading the company’s LNS when I decide v4 is prod ready.
If the form saving issue is blocking you from upgrading to v4, then please see my other comment (tldr: please look at the different tabs in the form, the validation error is probably in an other tab and is not related to your change).
As I mentioned in my previous reaction, I do remember that there was an issue with open file descriptors which was solved in v3, so an upgrade to v4 might not be required. Please see the changelogs and / or closed GitHub issues because I don’t remember when this was fixed, it was quite some time ago.
ok i will investigate further then…