Re: SS1+: real bad stability after woody-upgrade
christian mock wrote: > > On Mon, Aug 12, 2002 at 01:11:16PM +0200, Robert Waldner wrote: > > > >i'll be grateful for all hints that might lead to solving this. > > > > Yes, definitely. > > Well, another data point: my SS10 (2x 40 MHz supersparc) behaves badly > since about when woody became "stable" too -- it was actually running > woody before, but very solidly. > > In my case, it's the postfix "master" process which goes crazy -- the > last time it was hung in state "D", now it is in "R", both times > unkillable. funny thing is that I have problem with postfix (for about few months), cleanup is the program and the problem is that it just stops and waits, probably in kernel since it's unkillable, when I attach gdb and send interupt signal it is interrupted after MANY HOURS (it ends up in open function of standard c library). I have no idea why this happens, I haven't found any interesting messages in syslog (or on console). try to attach the gdb to master and see where it ends up (might take a long time, even days!) it's on intel machine though, kernel 2.4.18 erik
Re: SS1+: real bad stability after woody-upgrade
On Mon, Aug 12, 2002 at 01:11:16PM +0200, Robert Waldner wrote: > >i'll be grateful for all hints that might lead to solving this. > > Yes, definitely. Well, another data point: my SS10 (2x 40 MHz supersparc) behaves badly since about when woody became "stable" too -- it was actually running woody before, but very solidly. In my case, it's the postfix "master" process which goes crazy -- the last time it was hung in state "D", now it is in "R", both times unkillable. The console shows: Unsupported unaligned load/store trap for kernel at . Kernel panic: Wheee. Kernel does fpu/atomic unaligned load/store. Press L1-A to return to the boot prom f0029018 would be in srmmu_set_pte_nocache_viking according to System.map The kernel I'm running is kernel-image-2.2.19-sun4dm-smp, version 8, the very same kernel I'm running since months. My other sparc32, an IPC, is running well even with woody -- but "running" in that case means a very minimal install of woody (not even an MTA, just ssmtp) just idling along. That one has an uptime of 146 days and is running kernel-image-2.2.20-sun4cdm version 9. ciao, cm. -- ** christian mock in vienna, austria -- http://www.tahina.priv.at/~cm/ Wahrscheinlich ist die Klage billiger als eine Inseratenkampagne: "Wir haben das Internet nicht verstanden." -- Martin Hlustik ueber Ferrero in Sachen kinder.at
Re: SS1+: real bad stability after woody-upgrade
On Mon, 12 Aug 2002 18:50:30 +1000, Alexander Zangerl writes: >robert's ss1 with less mem is even worse, his highest uptime was <1d IIRC. Actually, my newest record since the upgrade is 13:07:20 up 3 days, 13:39, 1 user, load average: 0.28, 0.06, 0.02 I s/sendmail/exim/ to get that good, but even as the crashes aren't as freuqent as qith running sendmail, they're still biting every so often. I do log everything to the console, and log that to a file, but it simply drops dead, sometimes in the middle of a message, like: Aug 8 11:22:06 gonzales innfeed[245]: pizzaschleicher.snafu.priv.at:0 connected Aug 8 11:22:06 gonzales innfeed[245]: pizzaschleich >i'll be grateful for all hints that might lead to solving this. Yes, definitely. cheers, &rw -- -- "Die FPOe steht auf der Seite der deutschkaerntner Bevoelkerung" -- Martin Strutz, FPOe -- Was brauchts noch mehr? signature.ng Description: PGP signature
SS1+: real bad stability after woody-upgrade
as one of my friends already mentioned 1-2 weeks ago, there are apparently some very ugly problems with woody on very old sparcs. in my case it's a SparcStation 1+, in robert's a SparcStation 1. my box was running potato-stable and a bit of testing (mainly libc and trn) for more than a year without major gotchas. it was dead-slow but rock-solid. but since the upgrade to woody just weeks ago i've had three or four crashes already, and some of those were of the hard kind, i.e. the thing does not even respond to breaks on the serial console and needs to be powertoggled (lots of fun, the box is in colo at the other end of the world). robert's ss1 with less mem is even worse, his highest uptime was <1d IIRC. the system logs don't reveal anything; so far i have not tried to log stuff logged on the serial console, but that's next. the kernel has not changed, it's a self-built 2.2.19 since when that version came out. the box's feature-set has not changed either, mainly i'm running sendmail stripped down as much as i can (i.e. no ssl), inn2 and ssh. i don't really understand the problem. anybody else with an old box experiencing similar problems right now? ben, might recompiling the kernel with the compiler be a good idea? or maybe go to 2.2.20 or 21? (still i'm unsure how an unchanged kernel can suddenly show such lockups - and if it's not the kernel, what else can lock the box as badly as it does right now?) i'll be grateful for all hints that might lead to solving this. regards az -- + Alexander Zangerl + [EMAIL PROTECTED] + DSA 42BD645D + (RSA 5B586291) You know you're a Unix guy when your dreams start with #!/bin/sh. signature.ng Description: PGP signature