Operating System - Tru64 Unix
1748171 Members
4126 Online
108758 Solutions
New Discussion юеВ

Re: Tru64 5.1B PK4 multivolume clone

 
Ivan Ferreira
Honored Contributor

Tru64 5.1B PK4 multivolume clone

Hi all,

I'm having a problem with a multivolume clone. We have two EVA5000 with Continuous Access and severals virtual disks configured. Every virtual disk is a filesystem. One (and only one) of these filesystems is a multivolume filesystem:

# ls /etc/fdmns/hist04_dmn
dsk49c dsk52c

# ls /etc/fdmns/hist03_dmn
dsk16c
.
.
.
etc.

The hist03_dmn is a 500 GB domain, and hist04_dmn domain is 150 + 350 GB domain.

I create a clone for these filesystems to backup. The backup (with vdump) of all clone filesystems works, but i cannot backup the multivolume clone, vdump hungs when tries to get the directory and file count. (Sometimes i cannot kill the vdump process)

If i enter to the clone filesystem and run the ls command, it displays the files. If I run the ls -la command, the command hungs. On the others clone filesets the ls -la and vdump command works without problem.

It seems that when any command tries to get inode information, the command hungs.

Anobody seen something like this before?

I will move all the data to a single volume filesystem to backup, but i would like to know what happened here. I will keep the multivolume domain for troubleshooting.

Thanks in advance.
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
7 REPLIES 7
Venkatesh BL
Honored Contributor

Re: Tru64 5.1B PK4 multivolume clone

'freezefs' could help here. Check out the man page for more info.
Han Pilmeyer
Esteemed Contributor

Re: Tru64 5.1B PK4 multivolume clone

When you say "clone", what do you mean? Are we talking about an EVA snapclone (or snapshot) or a cloned AdvFS fileset?
Mark Poeschl_2
Honored Contributor

Re: Tru64 5.1B PK4 multivolume clone

How bad is the level of fragmentation in the domain you're having trouble with? Assuming you're doing AdvFS cloning and not the controller-based kind, cloning adds the overhead of doing extent map merges to the clone whenever the original volume is written to. I have seen this affect backups and system performance overall. What does 'defragment -vn' say about your multi-volume domain?
Ivan Ferreira
Honored Contributor

Re: Tru64 5.1B PK4 multivolume clone

I am using clonefset (advfs utilities) to make the clones.

The filesystem has been rebalanced to troubleshoot the problem, but it does not solve it.
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
Ivan Ferreira
Honored Contributor

Re: Tru64 5.1B PK4 multivolume clone

The output from the defragmento command is:

# defragment -nv hist04_dmn
defragment: Gathering data for domain 'hist04_dmn'
Current domain data:
Extents: 3864
Files w/extents: 20
Avg exts per file w/exts: 193.20
Aggregate I/O perf: 100%
Free space fragments: 222
<100K <1M <10M >10M
Free space: 0% 0% 0% 100%
Fragments: 17 156 2 47

It seems is not very fragmented.
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
Ivan Ferreira
Honored Contributor

Re: Tru64 5.1B PK4 multivolume clone

Also, at the time of the problem, vmstat showed free cpu and memory. iostat, cfsstat showed less activity on this filesystem than the others being backuped.

Thanks to all for your answers.
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
Han Pilmeyer
Esteemed Contributor

Re: Tru64 5.1B PK4 multivolume clone

When using cloned AdvFS filesets, freezefs is irrelevant.

You don't mention which version and patch kit you're running. There have been a lot of fixes to vfast, AdvFS migration/balancing/cloning in the latest patch kits. You may want to try again with the BL26 (PK5) patch kit for V5.1B.

If you want to find out what is wrong with the current version, then you may need to open a call with HP. A live dump of the system should help. You may get some idea of what's going on by attaching dbx to vdump and making one (or more) stack traces while it is hung.