Page 4 of 5

Re: So I had put accel-ppp live with heavy load

Posted: 25 May 2020, 08:20
by dimka88
Hi,

Just increase log level and set verbose for all modules, I think we can found the answer on accel-ppp detailed logs.
NOTE: Please, configure your firewall

Re: So I had put accel-ppp live with heavy load

Posted: 06 Jun 2020, 08:34
by hashbang
hi,
accel-ppp service restarted mysteriously.
Jun 06 00:15:10 hbbras01-153204 systemd[1]: accel-ppp.service: main process exited, code=killed, status=9/KILL
Jun 06 00:15:10 hbbras01-153204 systemd[1]: Stopped Accel-PPP.
Jun 06 00:15:10 hbbras01-153204 systemd[1]: Unit accel-ppp.service entered failed state.
Jun 06 00:15:10 hbbras01-153204 systemd[1]: accel-ppp.service failed.
Jun 06 00:15:10 hbbras01-153204 systemd[1]: Starting Accel-PPP...
Jun 06 00:15:10 hbbras01-153204 systemd[1]: Can't open PID file /var/run/accel-pppd.pid (yet?) after start: No such file or directory
Jun 06 00:15:10 hbbras01-153204 systemd[1]: Started Accel-PPP.
debug file is too big so extracting will take time. Above has been taken from message.
some nms graphs may reveal something @11:10* active users went down and after one hour a service restart
Screenshot from 2020-06-06 13-59-46.png
Screenshot from 2020-06-06 13-59-46.png (172.55 KiB) Viewed 13672 times
I'll post debug messages later on.
thanks

Re: So I had put accel-ppp live with heavy load

Posted: 06 Jun 2020, 08:42
by hashbang
yes service was busy as no info about active user during this period till it restarted itself
Screenshot from 2020-06-06 14-08-47.png
Screenshot from 2020-06-06 14-08-47.png (143.85 KiB) Viewed 13672 times

Re: So I had put accel-ppp live with heavy load

Posted: 06 Jun 2020, 15:08
by hashbang
no debug logs during this time
[2020-06-05 23:08:26.889] pppoe876: 0000000000511f87: send [LCP EchoRep id=a3 <magic 5a069ec4>]
[2020-06-05 23:08:26.890] ens5f1.630: recv [PPPoE PADI 84:35:2e:4e:0d:05 => ff:ff:ff:ff:ff:ff sid=0000 <Service-Name > <Host-Uniq 31420000>]
[2020-06-05 23:08:26.890] ens5f1.630: send [PPPoE PADO 00:1b:21:bc:02:26 => 84:35:2e:4e:0d:05 sid=0000 <AC-Name HBBras-153204> <Service-Name > <AC-Cookie f3b7aa403c923d0387050252b71e7f5e24e08e1f50a8643b> <Host-Uniq 31420000>]
[2020-06-05 23:08:26.892] ens5f1.630: recv [PPPoE PADR 84:35:2e:4e:0d:05 => 00:1b:21:bc:02:26 sid=0000 <Service-Name > <Host-Uniq 31420000> <AC-Cookie f3b7aa403c923d0387050252b71e7f5e24e08e1f50a8643b>]
[2020-06-05 23:08:26.908] ens5f1.124: 0000000000515562: fsm timeout 9
[2020-06-06 00:12:36.936] terminate, sig = 15

Re: So I had put accel-ppp live with heavy load

Posted: 07 Jun 2020, 20:38
by dimka88
Hi, can you run accel-pppd process in GDB?
https://accel-ppp.readthedocs.io/en/lat ... index.html

Re: So I had put accel-ppp live with heavy load

Posted: 06 Jul 2020, 10:17
by hashbang
Hi,
It has been confirmed that whenever there is a big network flap affecting 1k+ users, accel-ppp service restarts which aggravates the problem relogin so many users at once. I'm awaiting for new h/w to get debugging into gdb done.

thanks

Re: So I had put accel-ppp live with heavy load

Posted: 11 Jul 2020, 07:39
by hashbang
again network flapped same situation but this time accel-ppp service started every few minutes. I found after 10 minutes as my interim update is 10min. Restarted accel-ppp many times but problem didnt resolved. Rebooted the server then things went fine.

thanks and regards

Re: So I had put accel-ppp live with heavy load

Posted: 12 Jul 2020, 13:59
by dimka88
Hi, you can also try to add to [common] section parameter max-starting.

Code: Select all

max-starting=n
By default is disabled max-starting=0

Specifies maximum concurrent session attempts which server may processed.
But will be better to get coredump and capture logs

Re: So I had put accel-ppp live with heavy load

Posted: 13 Jul 2020, 14:08
by hashbang
ty

Re: So I had put accel-ppp live with heavy load

Posted: 18 Jul 2020, 12:11
by lbw
I'm running accel-ppp on multiple servers each doing 3Gbps load with mixture of PPPoE and L2TP without issue (touch wood) using 10-Gigabit X540-AT2, 4.9.0-11-amd64 and 8 x E3-1270 and accel-ppp has been running for about 150 days with ~1000 sessions. Can you scale horizontally?