I know, I’m bumping a year-old thread, but I have the same issue… with a slight twist: even when Cells is running after successfully booting, it still opens way too many open connections that do not get closed and remain in a sleep state ‘forever’.
The irony is that for months I have been trying to track down the
too many connections error to all sorts of other applications, with zero success. Every once in a while, the server I got will struggle to connect to MariaDB — and the WordPress installations I’ve also been running on the same server, as well as other apps, fail with this error. After all, they are not to blame…
When running the many tools to optimise or fine-tune MySQL there is always the mention that ‘too many connections are being aborted’ and the possible cause given for that is recurrent authentication failures. Again, for months I have been fearing that I was being the victim of a mysterious exploit that was attempting to guess MySQL passwords — although I could find nothing on the logs that could give me the slightest clue to the origin of such alleged attacks. I even tried to go very deep into the networking level to try to figure out what’s wrong with the packets sent and received by MariaDB. And again, this was another exercise in vain: there were no mysterious ‘attacks’ or failed authentication attempts coming from anywhere.
Tracking down where exactly in the Cells code database connections are being opened but not closed will be a nightmare; to make matters worse, at the time of writing, I’m having a slight issue compiling the current Cells code on either macOS or Linux — mostly because Cells originally didn’t use Go modules, but rather a popular, pre-modules tool — so I cannot even build a custom version of Cells to help with debugging. I’ve not given up (I used to successfully compile Cells without any problems whatsoever, on earlier Go versions, so I’m sure that, eventually, I’ll manage to do it again), but, in the meantime, I defer to the developers and the community to audit the code and figure out where exactly connections are being opened but never closed, being left in a sleep state.
It would be also interesting to see if more Cells users have encountered the same issue, and, if so, how they dealt with it. Currently, my best choice is to run a
cron job every now and then and clean up the sleeping threads; but that’s a crude hack, not addressing the core issue.
No wonder that my poor MariaDB server, although there are not that many processes running on the server, nevertheless consumes almost 5 GBytes (!!!) of real memory (not virtual!) — probably because it has to keep so many open connections that never get closed…