Hi Ajith,
Can you please get the fix done to registry and test it out. IIRC, the
deadlocking scenario in the code happens for update operations, the insert
operation sequence seemed correct.
Cheers,
Anjana.
On Mon, Feb 6, 2012 at 1:47 PM, Senaka Fernando wrote:
> Hi Amila, Anjana,
>
> DimuthuG
Hi Amila, Anjana,
DimuthuG did this sometime back and also produced a spreadsheet (should be
on google docs) on the sequences. But something that we probably missed (as
in this case) is the locking granularity. There is a subtle difference when
row-level and table-level locks are acquired in and n
Hi Amila,
On Mon, Feb 6, 2012 at 11:50 AM, Amila Suriarachchi wrote:
>
>
> On Mon, Feb 6, 2012 at 11:35 AM, Anjana Fernando wrote:
>
>> Hi Amila,
>>
>> On Mon, Feb 6, 2012 at 10:30 AM, Amila Suriarachchi wrote:
>>
>>>
>>>
>>> On Mon, Feb 6, 2012 at 12:50 AM, Senaka Fernando wrote:
>>>
Hi a
On Mon, Feb 6, 2012 at 11:35 AM, Anjana Fernando wrote:
> Hi Amila,
>
> On Mon, Feb 6, 2012 at 10:30 AM, Amila Suriarachchi wrote:
>
>>
>>
>> On Mon, Feb 6, 2012 at 12:50 AM, Senaka Fernando wrote:
>>
>>> Hi all,
>>>
>>> We discussed this offline, and the issue here is valid, Anjana can
>>> expl
Hi Amila,
On Mon, Feb 6, 2012 at 10:30 AM, Amila Suriarachchi wrote:
>
>
> On Mon, Feb 6, 2012 at 12:50 AM, Senaka Fernando wrote:
>
>> Hi all,
>>
>> We discussed this offline, and the issue here is valid, Anjana can
>> explain further if someone wants to understand it.
>
>
> Can you please do
On Mon, Feb 6, 2012 at 12:00 AM, Anjana Fernando wrote:
> Hi,
>
> I may have possibly found a solution for the registry deadlocking for
> concurrent puts/deletes. I was checking the SQL statements executed for a
> put, and the list is as follows.
>
> SELECT REG_PATH_ID FROM REG_PATH WHERE
> REG_P
On Mon, Feb 6, 2012 at 12:50 AM, Senaka Fernando wrote:
> Hi all,
>
> We discussed this offline, and the issue here is valid, Anjana can explain
> further if someone wants to understand it.
Can you please do that?
I also thought (as Senaka mentioned) update sequence should be consistent
across
Hi all,
We discussed this offline, and the issue here is valid, Anjana can explain
further if someone wants to understand it. But out of the three options,
right now, only #3 seems to be possible without breaking consistency. But
even this needs to be tested against all DBs, because we are momenta
Hi again,
And with regard to #3, if that's a safe option go for it. That's probably
due to somebody not realizing the possibility. Anyway, I'm not the expert
when it comes to SQL stuff, may be Sumedha can provide some insight?
Thanks,
Senaka.
On Mon, Feb 6, 2012 at 12:06 AM, Senaka Fernando wro
Hi Anjana,
So is this a complete sequence of operations? If so, once a write lock is
acquired, IIRC it will be held until the transaction completes - am I
missing something?
Also, there are some INSERT statements above the deletes. What about those?
Thanks,
Senaka.
On Mon, Feb 6, 2012 at 12:00
Hi,
I may have possibly found a solution for the registry deadlocking for
concurrent puts/deletes. I was checking the SQL statements executed for a
put, and the list is as follows.
SELECT REG_PATH_ID FROM REG_PATH WHERE
REG_PATH_VALUE='/_system/config/abc1' AND REG_TENANT_ID=0
SELECT REG_VERSION
11 matches
Mail list logo
