[opensuse-ppc] Few problems on P7 ppc
Hi all, as told earlier, we are running opensuse 13.1 ppc on a IBM P7 lpar ( thx to Dinar Valeev for the help), but we are running into a few problems. Is there someone who experiences the same problems, or maybe even better, solved them ;) We use our system for icinga monitoring of about 200 servers with a total of 5000 services. We migrated from x-series to p-series (P710) because the load that is generated by WMI checks for the windows systems in our environment. We have assigned max 2 processors and 4 gig of memory to our lpar (no swapping at the moment) Our problems are: - We tried to run MariaDB 5.5.33 on the system, which generated under some load (of ido2db for all our hosts), which generated a lot of "InnoDB: Assertion failure in thread 70366933971360 in file buf0buf.ic line 1277" errors. Moving to MySQL 5.5.6 reduced the number of errors, but certain queries are interpreted differently as the older versions of MYSQL (or this is a bug somewhere), and was not usable. We ended up moving our mysql-db back to x-series. - On all of our lnx systems we have AD users, implemented according to on http://en.gentoo-wiki.com/wiki/Active_Directory_Authentication_using_LDAP. When we enable the pam modules used in the AD authentication, su-ing to root gives a memory fault. - we have our monitoring over two systems at the moment, x-series with a few networkdevices (which will be transferred to p-series also, but we have a freeze in december). To have all the results in one interface, we installed check_mk on x-series and mk_livestatus on p-series. The mk_livestatus doesn't read a configuration setting (or doesn't interpreted it correctly), and gives a 0 for config_broker_options instead of the -1 in the config. Recompiling the file with the -1 hardcoded gives a working mk_livestatus. - mk_livestatus is used to read values from nagios using a commandpipe, but some of the values (always the same fields) are 0 instead of the unix timestamp they should have. Somehow this all seems a bit memory rmanagement related, but i am not a programmer / debugger... Has someone any ideas about these problems? Kernel version on the p-series: uname -a Linux <systemname> 3.11.6-4-default #1 SMP Wed Oct 30 18:04:56 UTC 2013 (e6d4a27) ppc64 ppc64 ppc64 GNU/Linux Hans Scheffers AIX / Linux Systeembeheer -- To unsubscribe, e-mail: opensuse-ppc+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-ppc+owner@opensuse.org
Noone experiencing the same problems, or part of them? Hans Scheffers AIX / Linux Systeembeheer ----------------------------------------
To: opensuse-ppc@opensuse.org Subject: Few problems on P7 ppc Date: Thu, 19 Dec 2013 10:41:00 +0100
Hi all,
as told earlier, we are running opensuse 13.1 ppc on a IBM P7 lpar ( thx to Dinar Valeev for the help), but we are running into a few problems. Is there someone who experiences the same problems, or maybe even better, solved them ;) We use our system for icinga monitoring of about 200 servers with a total of 5000 services. We migrated from x-series to p-series (P710) because the load that is generated by WMI checks for the windows systems in our environment. We have assigned max 2 processors and 4 gig of memory to our lpar (no swapping at the moment)
Our problems are: - We tried to run MariaDB 5.5.33 on the system, which generated under some load (of ido2db for all our hosts), which generated a lot of "InnoDB: Assertion failure in thread 70366933971360 in file buf0buf.ic line 1277" errors. Moving to MySQL 5.5.6 reduced the number of errors, but certain queries are interpreted differently as the older versions of MYSQL (or this is a bug somewhere), and was not usable. We ended up moving our mysql-db back to x-series.
- On all of our lnx systems we have AD users, implemented according to on http://en.gentoo-wiki.com/wiki/Active_Directory_Authentication_using_LDAP. When we enable the pam modules used in the AD authentication, su-ing to root gives a memory fault.
- we have our monitoring over two systems at the moment, x-series with a few networkdevices (which will be transferred to p-series also, but we have a freeze in december). To have all the results in one interface, we installed check_mk on x-series and mk_livestatus on p-series. The mk_livestatus doesn't read a configuration setting (or doesn't interpreted it correctly), and gives a 0 for config_broker_options instead of the -1 in the config. Recompiling the file with the -1 hardcoded gives a working mk_livestatus.
- mk_livestatus is used to read values from nagios using a commandpipe, but some of the values (always the same fields) are 0 instead of the unix timestamp they should have.
Somehow this all seems a bit memory rmanagement related, but i am not a programmer / debugger... Has someone any ideas about these problems?
Kernel version on the p-series: uname -a Linux <systemname> 3.11.6-4-default #1 SMP Wed Oct 30 18:04:56 UTC 2013 (e6d4a27) ppc64 ppc64 ppc64 GNU/Linux
Hans Scheffers AIX / Linux Systeembeheer -- To unsubscribe, e-mail: opensuse-ppc+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-ppc+owner@opensuse.org
On 09.01.2014, at 14:40, Hans Scheffers <hscheffers@outlook.com> wrote:
Noone experiencing the same problems, or part of them?
Unfortunately apparently not. But they do sounds heavily like endianness bugs. Could you please check whether you see the same issues when running in little endian mode using the (highly experimental!) Factory build for it? http://download.opensuse.org/ports/ppc/factory/iso/openSUSE-Factory-NET-ppc6... If they don't appear there we know at least what we have to look out for. Thanks! Alex -- To unsubscribe, e-mail: opensuse-ppc+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-ppc+owner@opensuse.org
On Thu, Jan 09, 2014 at 02:47:14PM +0100, Alexander Graf wrote:
On 09.01.2014, at 14:40, Hans Scheffers <hscheffers@outlook.com> wrote:
Noone experiencing the same problems, or part of them?
Unfortunately apparently not. But they do sounds heavily like endianness bugs.
Could you please check whether you see the same issues when running in little endian mode using the (highly experimental!) Factory build for it?
http://download.opensuse.org/ports/ppc/factory/iso/openSUSE-Factory-NET-ppc6...
If they don't appear there we know at least what we have to look out for.
Thanks!
This is even way way more bleeding edge than big endian ppc64 ... ;) Ciao, Marcus -- To unsubscribe, e-mail: opensuse-ppc+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-ppc+owner@opensuse.org
On 09.01.2014, at 14:50, Marcus Meissner <meissner@suse.de> wrote:
On Thu, Jan 09, 2014 at 02:47:14PM +0100, Alexander Graf wrote:
On 09.01.2014, at 14:40, Hans Scheffers <hscheffers@outlook.com> wrote:
Noone experiencing the same problems, or part of them?
Unfortunately apparently not. But they do sounds heavily like endianness bugs.
Could you please check whether you see the same issues when running in little endian mode using the (highly experimental!) Factory build for it?
http://download.opensuse.org/ports/ppc/factory/iso/openSUSE-Factory-NET-ppc6...
If they don't appear there we know at least what we have to look out for.
Thanks!
This is even way way more bleeding edge than big endian ppc64 ... ;)
Hell yeah, but it's a good way to boil down whether it's a "power" problem or an "endian" problem. Let's just cross our fingers it works at all ;) Alex -- To unsubscribe, e-mail: opensuse-ppc+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-ppc+owner@opensuse.org
Hi It has taken some time before we were able to rebuild a new system with a different starting image to see if the problems we had started with the used iso. I now have built the system using openSUSE-13.1-ppc64-NET-ppc64-Build0040-Media.iso. The problem with sudo is disappeared, so we can use active directory and sudo on the system. mysql hasn't been tested yet, I will try that this week. So my main problem seems to have changed... is there a way to 'update' the original system, so it uses the prekernel or so from this image instead of the original iso, or do i have to rebuild the complete system (running as a monitoring server with almost 7000 checks...) Grtz Hans Scheffers AIX / Linux Systeembeheer ----------------------------------------
Subject: Re: [opensuse-ppc] FW: Few problems on P7 ppc From: agraf@suse.de Date: Thu, 9 Jan 2014 14:51:56 +0100 CC: hscheffers@outlook.com; opensuse-ppc@opensuse.org To: meissner@suse.de
On 09.01.2014, at 14:50, Marcus Meissner <meissner@suse.de> wrote:
On Thu, Jan 09, 2014 at 02:47:14PM +0100, Alexander Graf wrote:
On 09.01.2014, at 14:40, Hans Scheffers <hscheffers@outlook.com> wrote:
Noone experiencing the same problems, or part of them?
Unfortunately apparently not. But they do sounds heavily like endianness bugs.
Could you please check whether you see the same issues when running in little endian mode using the (highly experimental!) Factory build for it?
http://download.opensuse.org/ports/ppc/factory/iso/openSUSE-Factory-NET-ppc6...
If they don't appear there we know at least what we have to look out for.
Thanks!
This is even way way more bleeding edge than big endian ppc64 ... ;)
Hell yeah, but it's a good way to boil down whether it's a "power" problem or an "endian" problem. Let's just cross our fingers it works at all ;)
Alex
-- To unsubscribe, e-mail: opensuse-ppc+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-ppc+owner@opensuse.org
-- To unsubscribe, e-mail: opensuse-ppc+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-ppc+owner@opensuse.org
participants (3)
-
Alexander Graf -
Hans Scheffers -
Marcus Meissner