This second terminal server is mostly identical except for the use of
Cyclades multiport cards rather than Comtrol.
On Sat, 15 Feb 1997, Jon Lewis wrote:
> Has anyone ever seen this sort of thing?
>
> Feb 15 09:17:55 endor pppd[3830]: pppd 2.2.0 started by riherd, uid 333
> Feb 15 09:17:55 endor pppd[3830]: Using interface ppp36
> Feb 15 09:17:55 endor pppd[3830]: Connect: ppp36 <--> /dev/ttyCh
> Feb 15 09:17:56 endor pppd[3830]: local IP address 205.229.48.4
> Feb 15 09:17:56 endor pppd[3830]: remote IP address 205.229.51.146
> Feb 15 09:17:56 endor pppd[3830]: local IP address 205.229.48.4
> Feb 15 09:17:56 endor pppd[3830]: remote IP address 205.229.51.146
>
> [this repeats _many_ times/s and then]
>
> Feb 15 09:18:10 endor pppd[3830]: local IP address 205.229.48.4
> Feb 15 09:18:10 endor pppd[3830]: remote IP address 205.229.51.146
> Feb 15 09:18:54 endor in.fingerd[4397]: connect from ewok.fdt.net
> Feb 15 09:18:46 endor kernel: Couldn't get a free page.....
> Feb 15 09:20:09 endor pppd[3776]: Hangup (SIGHUP)
> Feb 15 09:20:16 endor pppd[3776]: Modem hangup
> Feb 15 09:20:25 endor pppd[3776]: Connection terminated.
> Feb 15 09:20:28 endor pppd[3776]: Exit.
> Feb 15 09:18:10 endor pppd[3830]: local IP address 205.229.48.4
> Feb 15 09:18:10 endor pppd[3830]: remote IP address 205.229.51.146
>
> [more repetition clipped...then finally pppd seems to terminate]
>
> Feb 15 09:19:30 endor pppd[3830]: IPCP: timeout sending Config-Requests
>
> Note...it looks as if the system ran short on memory...I'm assuming the
> pppd process 3830 fell into some sort of "infinite" loop, terminated only
> when ipcp timed out, with a memory leak as a side affect. Look at the
> freelog (kept for historic reasons back when these boxes used to be pretty
> unstable)
>
> Here we are just before the problem. System is at half capacity portwise,
> plenty of RAM to spare, no load.
>
> 9:15am up 34 days, 19:08, 34 users, load average: 0.04, 0.07, 0.03
> total used free shared buffers cached
> Mem: 31012 30756 256 18700 2716 6572
> -/+ buffers: 21468 9544
> Swap: 61696 0 61696
>
> Here we are just after it...and that must have been a monster pppd process
> to have just released 17+mb of RAM...and look at that load avg!
>
> Sat Feb 15 09:20:30 EST 1997
> 9:20am up 34 days, 19:13, 34 users, load average: 96.34, 44.31, 16.85
> total used free shared buffers cached
> Mem: 31012 13528 17484 3012 196 1136
> -/+ buffers: 12196 18816
> Swap: 61696 9292 52404
>
> Shortly after this, the system returned to a normal state, and is still
> running. I suspect some of it is because I'm running an older syslogd
> that fsync's on every write...so that tight loop with the IP addresses
> being syslog'd must have really been a kick in the balls for the disk, not
> to mention all the swapping that happened. Time to upgrade some more
> systems.
>
> Kernel is 2.0.27, compiled ELF with gcc 2.7.2.1. pppd is a.out compiled
> with 2.6.3, libc 4.7.6. I've run a number of similar boxes for nearly 2
> years and never seen this happen.
>
> I just went looking for ppp-2.2.0g, and can't find it. It's not in the
> places I know to look, and ftp.sii.com looks as if Al may have left them.
> I'd hoped to check to see if this was perhaps a documented/fixed bug.
------------------------------------------------------------------
Jon Lewis <jlewis@fdt.net> | Unsolicited commercial e-mail will
Network Administrator | be proof-read for $199/hr.
________Finger jlewis@inorganic5.fdt.net for PGP public key_______