![]() | |
![]() |
| | Thread Tools | Display Modes |
#11
| |||
| |||
|
#12
| |||
| |||
|
#13
| |||
| |||
|
#14
| |||
| |||
|
|
-----Original Message----- From: info-ingres-bounces (AT) kettleriver...ting (DOT) com [mailto:info- ingres-bounces (AT) kettleriverconsulting (DOT) com] On Behalf Of Ingres Forums Sent: 12 August 2010 10:52 To: info-ingres (AT) kettleriverconsulting (DOT) com Subject: Re: [Info-Ingres] IMA query duration @Dejan Well... I just got back from a 3-week holiday, so I haven't processed all my e-mails yet, only the urgent ones. I remember seeing an e-mail wrt to my presentation. :-) I'll try to do that as soon as I can. I think it would be much better of course to have a trace point to log long queries (and long locks...)? Maybe I'll just have to join the code sprint next year. ;-) @Mason Wouldn't it be better to have a whole new trace point rather than risk breaking sc924? -- fba |
#15
| |||
| |||
|
|
Hi Gerhard I will send you the presentation via e-mail. BR Frédéric |
#16
| |||
| |||
|
|
Hi I have used a similar approach. An ESQLC-program with a loop that inserts the running queries in a table. I keep the queries running longer than a particular time interval in the table, so I can run reports on it later. I apply the same approach to get queries locking other queries longer than a specified time. I presented my approach on the IUA in London last June, so my presentation should be available on the IUA website. PostgreSQL has a parameter in the configuration file that says to log queries taking longer than a number of milliseconds (log_min_duration_statement). It would certainly be nice to have something like that in Ingres. It wouldn't probably be very difficult adding an parameter to the SC930 trace point (query recording) to specify a number of milliseconds. ;-) Best regards Frédéric |
#17
| |||
| |||
|
|
I originally raised this question, and then I coded some OpenROAD objects to do a similar thing to the ESQLC process mentioned below. The queries that have been there a while are stored in an array, they are written out to a table when its decided that they are slow number of seconds since first seen). If they are in LOCK state the blocking lock is found , using the usual ima method. It took a few hours to do but I already had all the ima sql setup, I just needed to manage the array of continued queries. |
![]() |
| Thread Tools | |
| Display Modes | |
| |