Has no one ever come across anything like this?
I have seen this issue in .28, .29 and now in .32.
It is 100% repeatable. Every time several checkouts are taking place
and an import is started (or vice/versa) all processes go to sleep, the
load drops and I/O is reduced to almost nothing...
-Ryan
On Wednesday, November 12, 2003, at 06:40 PM, Ryan Hunt wrote:
> And some more info...
> After running a svnadmin recover, future transactions again assume the
> run state rather than the sleep state.
>
> Would this indicate that importing simultaneous with a checkout will
> cause database corruption?
>
> -Ryan
>
> On Wednesday, November 12, 2003, at 06:00 PM, Ryan Hunt wrote:
>
>> So I have some more information on this problem.
>>
>> I have say 9 checkouts occurring simultaneously. Load is up to around
>> 10. Each process has about 9-10% CPU, and has a state of run. Files
>> are being written to the clients at about 30MB per min.
>>
>> As soon as I start an import for a 10th simultaneous svnserve process
>> all process assume the sleep state and drop down to 1% CPU or less.
>> Load drops to 1-2.
>>
>> The import is killed mid stream. All remaining svnserve processes
>> performing the checkouts remain in sleep state. While in sleep state
>> files are written to the clients at a rate that nonexistent.
>>
>> Any ideas on why this might occur and how to prevent it?
>>
>> -Ryan
>>
>> On Tuesday, November 11, 2003, at 05:35 PM, Ryan Hunt wrote:
>>
>>> I am running my repository on an HP-UX 11i system using the svn+ssh
>>> access method.
>>>
>>> When 2 separate users start 4 simultaneous checkouts each for a
>>> total of 8 checkouts, and a 3rd user initiates an import, the import
>>> grinds to a halt, and the checkouts proceed at a rate of .5MB per
>>> hour.
>>>
>>> Each transaction ranges in total size from 200MB to 2.5GB with the
>>> largest file size being 600MB.
>>>
>>> After all the transactions are canceled and another transaction is
>>> initiated the same slow pace is achieved.
>>>
>>> To revive the speed a reboot is required.
>>>
>>> Load never goes above 1.5, and no resources seem over taxed: disk
>>> I/O, memory, CPU all seem fine.
>>>
>>> Does anyone have any ideas or suggestions to allow multiple
>>> simultaneous large transactions??
>>>
>>> Ryan
>>>
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
>>> For additional commands, e-mail: users-help@subversion.tigris.org
>>>
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
>> For additional commands, e-mail: users-help@subversion.tigris.org
>>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Nov 13 21:16:36 2003