 | | | Strange kernel error on development RAC cluster (SLES9, 2 GB memor | Strange kernel error on development RAC cluster (SLES9, 2 GB memor 2007-07-26 - By Alexei_Roudnev
Back
I am just curious. I am running (set up 1 week ago for developers) RAC cluster
in development, on a couple of dell 670 workstation.
One of the systems have 2 GB only (other have 4) so I configured very small SGA
-s on it and balanced all so that system don't go to the swap.
It appear that system with lower memory get into the kernel error time to time,
which cause CSS to reboot it:
Jul 26 00:04:52 baikal2/baikal2 kernel: oracle: page allocation failure. order
:0, mode:0x20
Jul 26 00:04:54 baikal2/baikal2 kernel:
Jul 26 00:04:55 baikal2/baikal2 kernel: Call Trace:<IRQ> <ffffffff8016b9cf>{_
_alloc_pages+831} <ffffffff8016bb4d>{__get_free_pages+237}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffff8016f049>{cache_alloc
_refill+649} <ffffffff8016f4fe>{__kmalloc+94}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffff802eac61>{alloc_skb+81}
<ffffffffa00d518f>{:e1000:e1000_alloc_rx_buffers+127}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffffa00d485f>{:e1000:e1000
_clean_rx_irq+1743}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffffa00d3c68>{:e1000:e1000
_clean+1016} <ffffffff802ed414>{net_rx_action+132}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffff80142d03>{__do_softirq
+83} <ffffffff80142d95>{do_softirq+53}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffff80113dad>{do_IRQ+317}
<ffffffff80110ba7>{ret_from_intr+0}
Jul 26 00:04:55 baikal2/baikal2 kernel: <EOI> <ffffffff80171caf>{refill
_inactive_zone+383}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffff80171bb3>{refill
_inactive_zone+131} <ffffffff80172ea5>{shrink_zone+229}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffff80173816>{try_to_free
_pages+278} <ffffffff8016b8bf>{__alloc_pages+559}
Jul 26 00:04:55 baikal2/baikal2 kernel: <ffffffff8016bb4d>{__get_free
_pages+237} <ffffffff801a3c6a>{__pollwait+74}
Jul 26 00:04:55 baikal2/baikal2 logger: Oracle CSSD failure 134.
Jul 26 00:04:55 baikal2/baikal2 logger: Oracle CSSD failure 134.
Jul 26 00:04:56 baikal2/baikal2 kernel: <ffffffff802eb6c5>{datagram_poll
+21} <ffffffff802e51da>{sock_poll+26}
Jul 26 00:04:56 baikal2/baikal2 kernel: <ffffffff801a4008>{sys_poll+584}
<ffffffff801a3c20>{__pollwait+0}
Jul 26 00:04:56 baikal2/baikal2 kernel: <ffffffff801106b4>{system_call
+124}
Any idea, what it may be? Of course, RAC should not run 3 databases on 2 GB,
but technically all SGA and PGA fit and I have enough swap. I am not concerned
about this particular cluster, concern is that it may indicate overall system
/kernel bug.-- -- Original Message -- --
From: "Alexei_Roudnev" <Alexei_Roudnev@(protected)>
To: "Dimitrios Bougoulias" <D.Bougoulias@(protected)>; <suse-oracle@(protected)
.com>
Sent: Thursday, July 26, 2007 9:31 AM
Subject: Re: [suse-oracle] ASMLIB for IA64???
> You DON'T NEED ASM library for ASM. Just use raw devices and forget about
> ASMLib. ASMLib is just a piece of junk, doing nothing usefull.
>
> You must:
> - create raw devices using /etc/init.d/raw service or modified variation (I
> ise rawnames service);
> - start dbca, ask to create ASM, and assign these (one or more) raw disks to
> asm.
>
> When working with RAC, if root.sh can't initiate OCRFile (which must be a
> raw disk) or CSSFile, have a few tips:
> - sometimes it can't work with partitions,. sometimes with the whole disk.
> Looks as an oracle bug (10.2 only);
> - CSSFile must be owned by oracle, 'disk' group is not enough. To change
> ownership of the raw device, use
>
> But I'd better advice against SLES10 , Itanium and RAC together. Take SLES9
> SP3 _or_ take AMD/Intel server(s). This combination (SLES10 SP1, Itanium,
> RAC cluster) is more likely a big troublemaker because it is very uncommon.
>
> Here are my installations, but they are all for Intel and for SLES9 (I can't
> allow SLES10 for the database yet because of numerous problems with it).
>
>
> -- -- Original Message -- --
> From: "Dimitrios Bougoulias" <D.Bougoulias@(protected)>
> To: <suse-oracle@(protected)>
> Sent: Thursday, July 26, 2007 2:42 AM
> Subject: [suse-oracle] ASMLIB for IA64???
>
>
>> Hello,
>>
>>
>>
>>
>>
>> Does anybody know when ASMLib 2.0 will be available for ia64.
>>
>>
>>
>> (We need to migrate two Itanium servers to SLES10 with Oracle 10g R2 /
>> RAC -
>> Standard Edition)
>>
>>
>>
>>
>>
>> Thank you in advance,
>>
>>
>>
>> Dimitri Bougoulias
>>
>>
>
>
> --
> To unsubscribe, email: suse-oracle-unsubscribe@(protected)
> For additional commands, email: suse-oracle-help@(protected)
> Please see http://www.suse.com/oracle/ before posting
>
>
|
|
|