Upgaradation from 6.0.1 to 7.5 or 7.5A?

[vinaymanchinila]

Hi,
Right now we have our development environment at DS 6.0.1, we are planning to upgrade to either 7.5 or 7.5A.
1) Not sure which version 7.5 or 7.5A is better.
2) What kind of testing do I do with my current jobs after the upgradation and is there any testing to do before the upgradation to get a benchmark .

Thanks,

[kduke]

We just upgraded to 7.5.1 and are having lots of problems. I hear 7.5 is very stable. When we isolate the problems then we will fill you in.

I am curious. Has anyone downgraded from 7.5.1 to 7.5 and what were the issues?

[vinaymanchinila]

Hi Kim,
Can you let me know som eof the issues you are having which were not in 7.5.

[ml]

The best version will depend on your OS, you can check this in the Asc site.

After the upgradation, one of the first things you need to be sure is the compilation of your old job, and then you should test (run) a representative sample of your projects (jobs, routines).

You can read the readme file to see the DS7.5 improvements and take a look of the Inst and Upgrade Guide pdf (there you have a troubleshooting chapter)

[kduke]

The Designer seems to be unstable. If you pull up more than one Designer or too many jobs then it will error out and disappear. It does not leave jobs locked though. That helps. Also we cannot copy parts of a sequence in the Designer either in the same job or to another sequence.

The server seems to have weird errors too. We had some memory allocation error. We also have a memory leak. They said it was the Oracle 9i client but we were using the 8i client and reading from 8i. Hopefully we will resolve this soon.

[kduke]

I think our Designer problem is a Windows 2000 issue. The memory leak is a server side problem. I think a lot of issues are going away except for these 2. If you use XP and have no issues with 7.5.1 or 2000 and you do then please post.

The memory leak kicks in on about 2 million rows. We had this in 7.1 but now it seems like it effected more jobs. Maybe we just have more data.

[paulhill20]

[quote="kduke"]I think our Designer problem is a Windows 2000 issue. The memory leak is a server side problem. I think a lot of issues are going away except for these 2. If you use XP and have no issues with 7.5.1 or 2000 and you do then please post.

The memory leak kicks in on about 2 million rows. We had this in 7.1 but now it seems like it effected more jobs. Maybe we just have more data.[/quote]

I'm having issues as well; we tried 7.5 on AIX 5.2 and it seemed stable; we are now trying 7.5.1 on AIX 5.3 and I am having quite a time just trying to get "stable" executions of job streams; any suggestions you can provide would be appreciated.
Stephen de Vries

[kduke]

The memory leak appears to be an Oracle client issue. We have 2 versions of 8.1.7. The test and dev versions are the same and there is no leak. The production version is different and does not have the issue. The Oracle 9.2 version works as well. It only works with the Oracle 8 OCI plug in. We do not know why yet.

The Designer issue with sequences is in both XP and 2000. If you change several stages in a sequence then it blows up. Not sure what we are going to do. I want to go down to 7.5. Let you know soon.

[paulhill20]

[quote="kduke"]The memory leak appears to be an Oracle client issue. We have 2 versions of 8.1.7. The test and dev versions are the same and there is no leak. The production version is different and does not have the issue. The Oracle 9.2 version works as well. It only works with the Oracle 8 OCI plug in. We do not know why yet.

The Designer issue with sequences is in both XP and 2000. If you change several stages in a sequence then it blows up. Not sure what we are going to do. I want to go down to 7.5. Let you know soon.[/quote]

Out of curiousity, how did you determine there was a memory leak in the Oracle client - I'm seeing my memory returned/reused by the O/S. We are using 9.2.0.5.

I've had a number of strange issues:
- where a job hangs or dies when being executed from a sequencer,
- a job hangs on a lookup or a join (the child processes don't seem to report that they are done)
- a job gets a sigint or a sigkill with no further explanation

I'm attempting to use DS_PXDEBUG to find out what is going on. Do you have any suggestions as I know you've used the product for a decent amount of time as to how to "debug" these issues.

[kduke]

If you have a memory leak then you can watch the process size grow by using top or ps -ef. At some number of records it will be larger than the OS can support and will core dump. In UNIX a core file can be analyzed to find out why.