Unix Technical Forum

Re: [Info-Ingres] table lock escalations in errlog

This is a discussion on Re: [Info-Ingres] table lock escalations in errlog within the Ingres forums, part of the Database Server Software category; --> Hi Franky, > That looks promising. > > Will it work for several iidbms servers on 1 DBMS ? ...


Go Back   Unix Technical Forum > Database Server Software > Ingres

FAQ Members List Calendar Search Today's Posts Mark Forums Read
  #1 (permalink)  
Old 04-20-2008, 10:37 PM
martin.bowes@ctsu.ox.ac.uk
 
Posts: n/a
Default Re: [Info-Ingres] table lock escalations in errlog

Hi Franky,

> That looks promising.
>
> Will it work for several iidbms servers on 1 DBMS ?

Yes. You do not need to specify the trace point from a connection to
iidbdb on each server. Just the once from any server will do the trick.
And then its on for all the servers.

> And what about the remark that I found on the Ingres site: don't apply
> trace points on production systems ?


Thats seriously overstated.

Perhaps it should read don't absolutley rely on trace points.

Several trace points have been quite stable ever since Ingres was first
thought of and although the output format may have changed over time
their purpose is still much the same. Hands up anyone who wouldn't
use dm420, or qe90?

>
> I have hundreds of these lock escalations each day ( 1000's of tables,
> 100's of users, 10000's queries). Will it destabilize the system if I
> permanently apply this trace point ?


Not in my experience. I have a similar situation where a system is just
going nuts with 1500 odd lock escalations a day. I've been using sc924
to slowly winnow this down to a more reasonable number.

But the trace point can be easily set and then just as easily unset. What
I find useful is to target a specific time where the escalations are very
common, then use a cron job to turn on sc924 at a start time and then
use cron to turn it off at some other time.

What you may find is that the sc924 shows a cursor process like a
fetch as the query in question but it may not give you the cursor
declaration. Furthermore of the cursor names may be dynamically
generated and not easily tracked in the code. Its at that point you'll need
to use query_text auditing to get the details. And query_text auditing
can grind a system into the dust if not used judiciously.

So in short, I can personally recommend sc924 (Its the Ducks Guts, the
Bees Knees and I might even go as far as 'the Wasps Nipples') of a
trace point. But use caution if you have to drag out the query_text audit.

Marty
--
Random Duckman Quote #113:
Cornfed: You parked in a handicapped space.
Duckman: So what, no one ever notices except people who are
supposed to park there and Hell, I can outrun them anyday.

Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
Reply


Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On
Forum Jump


All times are GMT. The time now is 02:52 PM.


Powered by vBulletin® Version 3.6.5
Copyright ©2000 - 2009, Jelsoft Enterprises Ltd.
SEO by vBSEO 3.2.0
www.UnixAdminTalk.com