How long rollback

Then see which have been requested or mounted during the rollback and you could get an idea of how long it will take. Philip Sevetson. Dan, This reminds me of several discussions about different factors in rollbacks, inflight, and -START, over the years. I think we had one of the worst rollbacks in history measured by corporate bottom line cost at a place where I worked in or At the time, all of our archive logs went directly to tape.

It started tying up CICS which drove stock picks at the warehouses , and the system went down when we tried to force it out. We got through one archive log in four and a half hours of four or five logs in scope of the rollback ; our revised strategy was to move all of the needed archive logs to DASD while DB2 was down, and then to —START again. All of that is ancient, I know. But the question I have is this.

Both are unpredictable business events The DB2 restart is fortunately very unlikely in my experience of the last decade! Or will the volume be recalled like HSM does? Also, based on industry worst cases, what else do we need to worry about? So: is there a resource? RE: how to tell how long a rollback will take. Michael Hannan. Posted Jun 07, PM. This thread already has a best answer.

If a significant time was spent waiting for locks, or there were multiple transactions in that 4 hours, then the rollback will take less time. So it will give you info on a rollback in progress, but it will not help with the percentage complete of a generic query.

I believe backups and restores populate this, perhaps CheckDB. Not much else does. You must be logged in to reply to this topic. Or is this totally out of question :? May 01, - pm UTC. Extrapolation is a "maybe" idea. If the table has no indexes or the columns being modified are not indexed , extrapolation should be workable.

Indexes are the unknown variant here. How about a null taddr? Doug, June 12, - pm UTC. I'm trying to drop a schema that is "currently connected" even though all the sessions connected to it are marked for "KILL".

Why would that be? June 12, - pm UTC. I have a flag which shows a statement which is a huge select for update statement. The session, marked for kill, is holding about 13 different locks, many of them TOs on temp tables. If this transaction is not rolling back, what is it doing? An online index creation on the same table appears to blocked on an enqueue waiting for it.

June 28, - pm UTC. Well - as you mentioned - this situation has resolved itself. I took the nasty route of killing the dedicated server process to speed things along.

I know you hate that. The java app was doing absolutely nothing, and there were only 3 connections to the entire database. Mine, the one that was killed, and the one trying to build the index. I didn't think to look at session event.

What would you suspect? What about a session killed? So, how to follow the rollback make for this session? In addition, who executes this rollback? The server process dedicated or shared or a background process? Thanks Michel. July 21, - am UTC. Elapsed: Thanks for your help Michel. What affects the performance of rollback. Vladimir, March 15, - am UTC.

Hi Tom Thank you for providing information on how to determine how much time left to finish a transaction rollback. Is there any way to increase speed of rolling back changes? Can I increase the speed? Thank you for your time and effort.

March 15, - am UTC. This frees up locks held by the killed process immediately. What are these undocumented parameters? A reader, March 15, - pm UTC. It is 8. The problem is that the performance of the system degraded drastically while SMON rolls transaction back.

Yes, there are no locks. But all processes take longer than usual. So I think I am interested in the undocumented parameters that may help to finish rollback process as soon as possible Thank you. March 15, - pm UTC. Why do you do a ton of work make things slow to undo this ton of work make things slow more than once in a blue moon? Thank god it is a one time event. Very useful. Thank you. If you create new table create table as select based on old table you get a table with no partitions just a pk and rows.

I do not uderstand what the alter table exchange is going to do. I have tried to find an example of updating millions of rows in a partitioned table but I cannot find one. Are the steps I need to take as above? Where does the creation of local indexes come in? What I am asking for is an example of the above. April 14, - am UTC. Improve this answer. Brent Ozar Brent Ozar Sign up or log in Sign up using Google.

Sign up using Facebook. Sign up using Email and Password. Post as a guest Name. Email Required, but never shown. The Overflow Blog. Podcast Explaining the semiconductor shortage, and how it might end. Does ES6 make JavaScript frameworks obsolete? Featured on Meta.