cancel
Showing results for 
Search instead for 
Did you mean: 

Backup failed with 83/84

Smalling
Level 3
Hi All,

I need a help , in my environment I see backup of set of clients failing with sc:83/84

I see the below errors inside the job details
Most of the time memory allocation failed or sts open failed .

But restart always helps the backup to complete .

Master 773 HPux
Media 773 HPux
Clients : win2k8
Target :msdp
Dedupe : always use media server.

I have 24 GB of memory in both media and client.

Need some direction to go about it.

Issue is on going , not came up in the recent times ..
5 REPLIES 5

Marianne
Moderator
Moderator
Partner    VIP    Accredited Certified
Please show us Details tab of one or more failed jobs.

What is the size of the MSDP pool?
We have seen in practice that NBU needs about 1.5GB memory for each TB of MSDP storage.

Job Sequence :

11/21/2016 16:43:17 - Critical bptm (pid=14991) image open failed: error 2060002: memory allocation

11/21/2016 16:43:17 - Info bptm (pid=14991) EXITING with status 83 <----------
11/21/2016 16:43:17 - Info Servername (pid=14991) StorageServer=PureDisk:XXXXXXXXXX; Report=PDDO Stats for (Servername): scanned: 1 KB, CR sent: 0 KB, CR sent over FC: 0 KB, dedup: 100.0%, cache disabled
11/21/2016 16:43:18 - Info bpbkar (pid=4248) done. status: 83: media open error
11/21/2016 16:43:18 - end writing
media open error (83)

 

bptm:

04:13:20.362 [15006] <16> 11714733:bptm:15006:Servername: [ERROR] PDSTS: impl_write_image: exit (0 0x1 2060017:system call failed)
04:13:20.362 [15006] <16> 11714733:bptm:15006:Servername: [ERROR] PDSTS: pi_write_image_v7: impl_write_image() failed (2060017:system call failed)
04:13:26.953 [15049] <2> dsml_extract_diskvolumeinfo: dsmlib_popext.cpp: 550 sev 2 diskvolumeinfo.system_tag = Unknown
04:13:29.351 [15060] <2> dsml_extract_diskvolumeinfo: dsmlib_popext.cpp: 550 sev 2 diskvolumeinfo.system_tag = Unknown

 MSDP SIZE : 14 TB

I don't think you have enough memory on that media server.  Have you opened a case on this?  Support can validate 100% whether you do or don't based on process usage and what's going on with spoold, etc., but as Marianne said, rule of thumb is still roughly 1.5GB per TB of MSDP space.  That means you need 21GB of RAM available for MSDP related operations, and on a 24GB system, you're likely not getting that.  My familiarity with HP-UX memory management and OS needs isn't what it used to be, but I know if this were a Linux system you'd be shaving things mighty close and I'd be telling you to add RAM.

 

Charles
VCS, NBU & Appliances

Hi Chas,

I am constantly noticing that free memory is @ 700MB OR near that point on the media server .
I am also noticing that spools utilising 45-47 ℅ of memory even in non backup hours.

Is there any way that I can prove that current memory is insufficient other than our job details?

The fact that you regularly have only 700MB of free memory on a 24GB system AND you're failing your jobs isn't enough to tell you that you have insufficient RAM to do the work you're asking the system to do?

I'm not really sure how to answer the question.  There's not a log that says "you have insufficient memory".  You aren't built to the best practices for an MSDP server of the size you're trying to run, IMO.  As has been said, 1.5GB per TB of MSDP -- that's JUST for MSDP, then you've got your OS, the rest of NBU, etc. to cram onto your box.

If you're looking to do PID follow-through and strace work to see which process is being starved specifically or which is starving the system, that's more detailed an effort than a discussion forum should be used to do, and you should have good knowledge of the process flow of NBU before you start diving too deep into it.  That's why I suggested logging a support call.  If you're not convinced on the face of things with the known facts and need some deeper root cause/silver bullet to point to, support can help get to that point.  

Charles
VCS, NBU & Appliances