Skip to main content

!Hubzilla Support Forum !Hubzilla Development

This was very hard to debug: The channel of @Olaf Wendt (Hz) was unexpectingly extremely slow:

time curl >/dev/null
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  352k    0  352k    0     0   2963      0 --:--:--  0:02:01 --:--:-- 82067

real    2m1,729s
user    0m0,076s
sys     0m0,027s

Well, 120 seconds is PHP's timeout value, so something had to timeout causing this problem. I finally found the time to debug the problem: I installed tideways to trace the calls and found the culprit: Hubzilla tried to connect to @Stefan Münz former hub which has been disabled, causing the aforementioned timeout (look at the callgraph attached).
To test ist, I added something to /etc/hosts:

And Olaf's channel became fast again :-)

I'm wondering how to prevent this... and without a debugging tool like Tideways (thank God I knew this tool from my former employee) it's next to impossible to find the problem

@Mark Nowiasz In my understanding, fetching data from other hubs/pods/nodes should be done by "worker" processes independant of the UI, started by a cron job. If such a job runs out of time for one or two minutes because of waiting for a server response, it doesn't matter.