High read io from w00x (SMCO) processes?
From: <"">
Date: Tue, 22 Dec 2020 20:38:16 +0000
Message-ID: <SJ0PR08MB66071F2D1F9AEF6AA32F20B1FEDF0_at_SJ0PR08MB6607.namprd08.prod.outlook.com>
Our storage group contacted me this week complaining that one of our database servers had started experiencing a much higher i/o load than previously (as of last Friday). Looking at the database in question did not show especially high activity.
Date: Tue, 22 Dec 2020 20:38:16 +0000
Message-ID: <SJ0PR08MB66071F2D1F9AEF6AA32F20B1FEDF0_at_SJ0PR08MB6607.namprd08.prod.outlook.com>
Our storage group contacted me this week complaining that one of our database servers had started experiencing a much higher i/o load than previously (as of last Friday). Looking at the database in question did not show especially high activity.
pidstat -dl showed that the vast majority of i/o activity is coming from w00x processes and is from reads rather than writes.
A few things that might be useful:
Oracle version 12.1.0.2.0
The database uses NCLOBs/securefiles
_enable_space_preallocation=3
There was unusually high DB activity on the database which touched the NCLOBs on the Friday when the issue started. But the high i/o issue continued over the weekend (when DB activity was very low) and all day Monday and Tuesday (high DB activity during the day, low DB activity during the night).
So my questions are
- Is it possible that SMCO is engaged in some very long running cleanup from the DB activity on Friday? Could is somehow be stuck on something?
- Is there any way to trace what work the w00x processes are doing?
- Any merit to the tempting notion of setting _enable_space_preallocation=0, waiting a bit, and then setting it back to 3?
- Any other notions?
As always, thanks in advance for any advice/suggestions.
Best,
Jay
-- http://www.freelists.org/webpage/oracle-lReceived on Tue Dec 22 2020 - 21:38:16 CET