Known issues upgrading to NW 7.4

Have no personal experience with 7.4 but they do have a doc. with known issues:

https://powerlink.emc.com/nsepn/webapps/btg548664833igtcuup4826/km/live1/en_US/Offering_Technical/Technical_Documentation/300-004-420_a01_elccnt_0.pdf?mtcs=ZXZlbnRUeXBlPUttQ2xpY2tDb250ZW50RXZlbnQsZG9jdW1lbnRJZD0wOTAxNDA2NjgwMjZjODI0LGRvY3VtZW50VHlwZT1wZGYsbmF2ZU5vZGU9MGIwMTQwNjY4MDI2YjU4MA__

Vic

We are running several installs of 7.3.2 and the Java GUI takes some getting used to. 7.4? You're more brave than me. Let me know how it goes.

Thanks,
Dennis

Thanks, Vic.

I will read the release notes for 7.4

Martijn

I totally agree with everybody that upgrading to 7.3.2 is a saver bet, but the powers that be have decided that we should upgrade to 7.4. I might be able to change their minds if I have compelling arguments not to go to 7.4. We have 'garantees' of EMC that this is going to work. However we recently attempted an upgrade to 7.3.2 Jumbo patch (build 386), which went wrong and we had to revert back to 7.2.2 (thank God for BCV's). So I can only imagine what might happen if we upgrade to NW 7.4

Do you have a confirmation about what went wrong when upgrading to 7.3.2? If it's a known problem, you can just verify if this was fixed on 7.4. If not fixed, then you'll have arguments to not do this.

Maybe you could share with us the problems you had when upgrading to 7.3.2, then we can try to verify if it is a known problem fixed on newer releases, or some bug that might still exists.

Another thing that you need to verify is if you use daemon.log to generate reports or other info. EMC has changed the way logs are handled in 7.4, you need one tool to get it in human readble format.

For the ones that have their own report tools, some effort will be needed in order to get that working on this new version.

We upgraded to 7.32 on solaris sparc, had a lot of problems, and upgraded to 7.4 - it has been miserable. We are still trying to get consistent savegroups to run, and savegroup notifications to work properly. If your current version works, I'd wait. I don't think we can roll back to pre 7.3, or I would have.

I greatly regret the upgrade, it has been very costly. We've put in a recent hotfix, and it kinda helped, but things are still poor.

Daniel, it's just one line to dump it in old format. Actually, if you check new logs resource you will find entry where you can force it to be written in parallel with raw log too.

skn, please give us your horror stories

Well, this week it's not feeling as horrible so far, but there have been a few -

A complicating factor is I don't believe inquire or jbverify work properly with 7.4, solaris 10 sparc, and IBM LTO-3 drives. Release notes mention this for a similar setup with IBM LTO-2 drives, so I assume it's the same problem. I should have read the release notes more closely (I saw LTO-2 and moved on..), I guess. I haven't heard from support if this is true for my setup, but I am assuming it is. This issue does not hurt anything, but led me to suspect problems with my jukebox, and waste a lot of time.

With 7.3 and 7.4 non-hotfixed installs, when I ran large groups they would sometimes abort with no explanation early in the backup process. These were usually for incremental backups, sometimes if I ran from a command line with -v I would see a segmentation fault, othertimes I wouldn't. Many clients wouldn't be started. Most of my testing was done with small groups of 4 or 5 clients at a time, which worked fine.

With 7.4 pre-hotfix, notification messages got truncated at about 500K. So details of failed savesets are lost.

Now with the hotfix, messages are not truncated. And so far, no mysterious group failures.

Still, notifications don't always get sent. Sometimes a group appears to be done, but never completes or aborts, so no message is sent. Other times a group does complete, but a message is sent. For small groups, I get a message. This throws a wrench is reporting systems that rely on notfications. You can use the gui to get by.

And finally, the hotfix is just some binaries with no explanation of issues possibly fixed, and I see nowhere to get any information on them. I just heard 'hot fix solves that issue' was given the url.

In hindsight, I'd wait for official patches to 7.4 for solaris sparc. I think 7.3 was more like version 8.00.

jbverify is not supported under Solaris 10. For inquire, depending on what you use as hba driver, inquire might have difficulties as well. This was very well documented in release notes for 7.3.x (I guess same applies to 7.4).


Wow... I never saw that before (at least not with 7.3.2 which I use in production). I guess you already did that, but just to check - you did disable nsrauth?


I assume you could still see that in daemon.log, right? Do you have LGTsc reference number for this problem?


Do you use native NW notifications or you have something on your own? I wonder if you would change notification action to be written to disk and then picked up with mailx or something you use would that change anything. I guess it wouldn't if that would be core notification issue...


What binaries are inside the patch?


Yes, EMC changed naming convention for some reason and confused people around the globe with that.

No, I did not. I've got an open case with support and have just been following their advice (so far, just run the hotfix is all I've heard). I did not see any authentication errors when we had this issue though.



Yes, it showed up properly in daemon.raw -and via the gui. It just got chopped off on notificaitons. I found a very old bug from version 5 of networker, that referenced a problem with networker's printf vs solaris printf. I assumed it was the old bug reappearing.

They referred me to - LGTsc07256

I'm not sure where to find information on that reference number. Powerlink returns nothing in search for it.



We use the native notifications. Since I started having problems I made multiple actions to be sure, one writes to disk, another just does the usual email. Our normal routine is to dump it to disk, process, then send a variety of email. I simplified it to the basics when started having problems.



nsrd and savegrp

Well, in 7.3.x if you would disable nsrauth you would resolve 80% of problems. And description like many groups running or client and then never finishing smells like it. Ask support what do they think of. We did discuss steps on how to do it here on several threads so you may wish to do it yourself too.


No, it is fresh new bug id... on PL you will find it once they create KB article for it or get it fixed in next maintenance release.

When upgrading to 7.3.2 on Windows there wasn't just one thing that made us decide to go back to 7.2.2. It was all kinds of strange errors that compounded one on to the other. What took a lot of time was the database conversion. We did not take into account that it took several hours for this to complete. After the installation itself, which went kind of okay it was time to test the new software. Firstly we wanted to test an RMAN backup that had to be started by an database admin. Somehow he messed up the script and he wanted to recover that script. On the unix client that could not start nwrecover. So this was like the chicken and the egg problem. We could not test RMAN backups and we could not recover the RMAN scirpt. Sorry if this all a little fague, but I am new to the Networker backup world and I didn't perform the upgrade myself. Furthermore there was no report made afterwards on what went wrong so it is hard to recall. I think in the end we got constricted by time. We may have stayed with the new 7.3.2 version if we had more time to solve the problems that we had.