[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-ha
Subject:    Re: [Linux-HA] Re: Linux-HA Digest, Vol 24, Issue 45
From:       Alan Robertson <alanr () unix ! sh>
Date:       2005-11-28 4:10:04
Message-ID: 438A831C.5070909 () unix ! sh
[Download RAW message or body]

Andrew Beekhof wrote:
> On 11/23/05, Alan Robertson <alanr@unix.sh> wrote:
>> Andrew Beekhof wrote:
>>> On 11/22/05, Valentin Eberhardt <veberhardt@amadeus.com> wrote:
>>>>> ----- Message from Andrew Beekhof <beekhof@gmail.com> on Tue, 22 Nov 2005
>>>> 12:12:30 +1100 -----
>>>>> To:    General Linux-HA mailing list <linux-ha@lists.linux-ha.org>
>>>>> Subject:       Re: [Linux-HA] Re: Linux-HA Digest, Vol 24, Issue 45
>>>>> On 11/11/05, Valentin Eberhardt <veberhardt@amadeus.com> wrote:
>>>>>> Hi!
>>>>>>
>>>>>> I run heartbeat-2.0.1-1 on a 3node linux 2.6 cluster having configured 2
>>>>>> resource groups (I've attached the cibadmin -Q to this email)
>>>>>>
>>>>>> Two things I experience:
>>>>>>
>>>>>> 1. is that the tengine process goes sometimes probably in to a loop using
>>>> as
>>>>>> much CPU as it gets.
>>>>>>
>>>>> can you send the TE logs from just before until just after this period
>>>> please?
>>>>
>>>> The TE is endlessly  looping, i.e. it's not transient.
>>> i'm trying to get a sense of what it might be doing in that loop.
>>> unfortunately the strace information just shows it polling for input
>>> which is quite normal.
>> The return value of zero from poll indicates that it's not receiving
>> anything.
>>
>> It's very odd that we're only polling one file descriptor.  That means
>> we're in one of the corner cases in the IPC code - because normally we
>> have several sources that we're polling each major iteration.
>>
>> Either line 729 or line 1032 in ipcsocket.c.  Line 1032 is obviously in
>> the middle of a loop - but I don't think that this loop will do a send()
>> call.
>>
>> The caller is either calling socket_waitin() or socket_waitout().
>> Socket_waitin() waits until a packet is eventually received, or
>> socket_waitout() waits until a packet is eventually sent.
>>
>> Since the writes on file descriptor 4 are succeeding, it's not very
>> likely that the client on the other side has gone away.
>>
> oh ok - so its definitely a messaging layer problem and nothing i've done?
> (thats what i suspected but its nice to be sure)

I'm not so sure as that.  The fact that you're writing every time around 
the loop is very odd to say the least.

-- 
     Alan Robertson <alanr@unix.sh>

"Openness is the foundation and preservative of friendship...  Let me 
claim from you at all times your undisguised opinions." - William 
Wilberforce
_______________________________________________
Linux-HA mailing list
Linux-HA@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
[prev in list] [next in list] [prev in thread] [next in thread]