You are viewing limited content. For full access, please sign in.

Question

Question

Oracle upgrade

asked on January 27, 2014

Upgrade to RIO Production environment was started on Thursday - RIO Oracle 9.0 to 9.1.

The upgrade seemed to hang but this upgrade can take up to an hour or more.

After waiting 16 hours, LF Support then suggested an index to improve performance, however once applied to the Test environment (which also
continues to hang for one repository), no progress has been apparent.

Meanwhile, the DBA noted that there seemed to be inefficient code that was causing full table scans for every step.

The advice provided was to wait, but the upgrade has been running now (ie Sunday at 7:00 pm) for over 72 hours.

Have any VARs had experience with an Oracle upgrade?

2 0

Answer

APPROVED ANSWER
replied on January 27, 2014

There's a defect in the 9.0 -> 9.1 upgrade script that causes upgrades to never complete in some circumstances. It will be fixed in LFS 9.1.1, which is going to be released in the next couple of weeks. Oracle Database users should hold off on upgrading to 9.1 until 9.1.1 is out. We apologize for the trouble. If you need to upgrade now, LF technical support can furnish a patch.

3 0

Replies

replied on January 27, 2014

This sounds exactly like what my DBA tells me every time we perform an upgrade or other database-focused task in Laserfiche on Oracle. He is constantly running into full table scans and tries applying indexes to fix it.

 

If Laserfiche can't figure out what indexes to apply, see if you DBA is willing to dig deeper and try applying indexes. There is a way to do this on the fly and monitor the progress of the SQL statements to see if it helps. 

 

Let me know what you find out. We will be upgrading from 8.3 to 9.1 in a few months. I'd love to avoid this problem if possible.

0 0
replied on January 28, 2014

It's not always a matter of having indexes.

I have tuned a few queries (4-5) in the past, and fortunately many are now in stored procedures so you can tweak them.

I have not created any index so far, but changed the statement syntax for instance, replacing NOT IN with EXISTS and it made an immense difference (few seconds compared to many minutes).

Some of that was implemented by the LF team in 9.0.

But of course during an upgrade it is even more difficult since you don't know beforehand what will happen, and then you cannot try again as these large updates are only meant to be run once. That's why I always do that in test first whenever possible.

When it's slow in production you can look for ways to improve it  as you can do multiple tests (searching for a document, or any specific task that's consistenly slow).

 

Most definitely the LF team should have a large datastore to test with, not only a few hundred documents. They would uncover problems before they get to customers.

 

Upgrading from 8.3 to 9.0 was fine, btw. The only big problem was to get into the properties screen of a volume. That can take 30 minutes to display.

0 0
replied on January 28, 2014

Most definitely the LF team should have a large datastore to test with, not only a few hundred documents. They would uncover problems before they get to customers

You give them more credit than I do. I just assume they don't test the Oracle upgrade scripts at all. wink I've run into a few issues when upgrading in the past that would fail no matter how many documents are in the repository.

 

Upgrading from 8.3 to 9.0 was fine, btw. The only big problem was to get into the properties screen of a volume. That can take 30 minutes to display.

It takes LONGER in 9.x? I thought a few minutes was too long. Thanks for the heads up. Glad to hear you didn't have much of an issue with the upgrade. I've been concerned because of all the database changes made for business processes.

 

0 0
You are not allowed to follow up in this post.

Sign in to reply to this post.