PPPoE keepalive

PPPoE related questions
Post Reply
hugleo
Posts: 59
Joined: 13 Apr 2016, 14:28

PPPoE keepalive

Post by hugleo »

Today I saw in the server hundreds of connections that have terminated (20% of total) and reconnected. All those peers ended the connection with the message: <Acct-Terminate-Cause User-Request>]
I believe it may have been that the pppoe client peer sent LCP echo messages and accel-ppp did not answer. So the client has sent a message of LCP termination request (TERMREQ) to end the session.
Is there some accel-ppp config setting to help stabilize/keepalive the sessions a little more?
Dmitry
Администратор
Posts: 954
Joined: 09 Oct 2014, 10:06

Re: PPPoE keepalive

Post by Dmitry »

i suggest to analyze logs first to find exact reason
hugleo
Posts: 59
Joined: 13 Apr 2016, 14:28

Re: PPPoE keepalive

Post by hugleo »

Dmitry wrote:i suggest to analyze logs first to find exact reason
I analyzed some logs but I can't find yet the reason:

Follow atached the /var/log/accel-ppp/accel-ppp.log in the link:
http://www.sametext.com/view/raw/0721ca7d

The problem start after the [2016-04-14 14:40:36]
[2016-04-14 14:40:36]: info: ppp1394: send [RADIUS(1) Accounting-Request id=94 <User-Name "uuuuuu"> <NAS-IP-Address hh.hh.hh.hh> <Acct-Terminate-Cause User-Request>]
...
Tons of lines like that.

dmesg
no new lines

/var/log/messages
no new lines

/var/log/accel-ppp/emerg.log
no new lines
Dmitry
Администратор
Posts: 954
Joined: 09 Oct 2014, 10:06

Re: PPPoE keepalive

Post by Dmitry »

i see tons of Acct-Terminate-Cause Lost-Carrier
this message is lcp echo related too
so what i can suggest is increase lcp-echo-failure value or lcp-echo-timeout
hugleo
Posts: 59
Joined: 13 Apr 2016, 14:28

Re: PPPoE keepalive

Post by hugleo »

I've been checking some clients and I've found that Acct-Terminate-Cause Lost-Carrier are the pppoe clients that doesn't support send the LCP termination request (differents devices).
So are duplicated very old connections due the fact that clients already start new one. I'll enable single-session=replace for it...

That problem not harm too much since it's happens only one time since production (two weeks) :)
Post Reply