don't wait rlm_sql response
Hi all, When using rlm_sql, it always wait for response from database. Is there any way to configure to avoid it? (We don't wanna wait for its result after prosessing) accounting { detail sql <-wish to prosess it here without waiting for result } We'd appreciate if anyone can answer this promptly. Thank you,
Mitsuhiro Nakamura wrote:
When using rlm_sql, it always wait for response from database.
That's how it works.
Is there any way to configure to avoid it? (We don't wanna wait for its result after prosessing)
Why?
accounting { detail sql <-wish to prosess it here without waiting for result }
We'd appreciate if anyone can answer this promptly.
That's nice, but no one is getting paid to help you. Asking for "prompt" responses is not helpful. Alan DeKok.
Alan, Thank you. Our database has some problem and the response is slow. Since the problem itself seems to take long time to fix it, we wanted to avoid it. If there is no way to do, then we will try to find the otherway. Regards, Nakamura
Mitsuhiro Nakamura wrote:
When using rlm_sql, it always wait for response from database.
That's how it works.
Is there any way to configure to avoid it? (We don't wanna wait for its result after prosessing)
Why?
accounting { detail sql <-wish to prosess it here without waiting for result }
We'd appreciate if anyone can answer this promptly.
That's nice, but no one is getting paid to help you. Asking for "prompt" responses is not helpful.
Alan DeKok.
On 2 avr. 2013, at 04:13, Mitsuhiro Nakamura <nakamura@1pacific.ddo.jp> wrote:
Alan,
Thank you. Our database has some problem and the response is slow. Since the problem itself seems to take long time to fix it, we wanted to avoid it.
Look at the decoupled accounting virtual server example. Olivier
On Tue, Apr 2, 2013 at 9:13 AM, Mitsuhiro Nakamura <nakamura@1pacific.ddo.jp
wrote:
Alan,
Thank you. Our database has some problem and the response is slow. Since the problem itself seems to take long time to fix it, we wanted to avoid it.
Are you using sql ONLY for accounting? If yes, then Olivier's suggestion on using decoupled accounting might help. In fact, on some scenarios where the db is actually fast but has a somewhat high latency (e.g. mysql cluster), I'd highly recommend that. However, since you say the db response is slow, you'd only be putting band-aid on the wound. The detail file reader may very likely end up hours or days behind the actual accounting packet. If there is no way to do, then we will try to find the otherway.
Fix the db. It's the right thing to do. The usual causes are: - too much data (e.g. tens of millions of rows from years of accounting data in a single table) - inefficient indices (too many index which slows down writes, or no index used which slows down reads) - not enough disk IOPS available (e.g. using 2 x HDD in mirror mode for a db with hundreds of writes requirement per second) - default, untuned db A qualified dba should be able to help you. -- Fajar
participants (4)
-
Alan DeKok -
Fajar A. Nugraha -
Mitsuhiro Nakamura -
Olivier Beytrison