AIX dump device not showing accurate size

I am trying to configure dump devices on my AIX server. Running 7100-03-04-1441. My dump device needs to be about 2GB in size. My PP Size is 1024MB, so I create the device with 2 PPs. When I run lslv on the dump device, it shows the 2 PPs, and a PP Size of 1024 megabytes. However, a dumpcheck -p shows that my dump device is only 256MB in size. dumpcheck seems to think my PP Size is only 128MB. Here you can see the output of lslv dump2lv showing the PP Size and PPs:

LOGICAL VOLUME:     dump2lv                VOLUME GROUP:   rootvg
LV IDENTIFIER:      00c1077000004b0000000166fa6a2f10.11 PERMISSION:     read/write
VG STATE:           active/complete        LV STATE:       opened/syncd
TYPE:               sysdump                WRITE VERIFY:   off
MAX LPs:            512                    PP SIZE:        1024 megabyte(s)
COPIES:             1                      SCHED POLICY:   parallel
LPs:                2                      PPs:            2
STALE PPs:          0                      BB POLICY:      non-relocatable
INTER-POLICY:       minimum                RELOCATABLE:    yes
INTRA-POLICY:       middle                 UPPER BOUND:    8
MOUNT POINT:        N/A                    LABEL:          None
MIRROR WRITE CONSISTENCY: on/ACTIVE
EACH LP COPY ON A SEPARATE PV ?: yes
Serialize IO ?:     NO
INFINITE RETRY:     no                     PREFERRED READ: 0

And here you can see dumpcheck showing the size being only 262144 kb:

# /usr/lib/ras/dumpcheck -p
The largest dump device is too small.

Largest dump device
         dump2lv
Largest dump device size in kb
         262144
Current estimated dump size in kb
         1955880

My rootvg does show that the size used is indeed 2GB. So where is this extra space going, and why is dumpcheck not reporting all the size for my dump device?

Thanks in advance!

Just because you create a logical volume of type "dump" doesn't mean it is used as dump. Use the sysdumpdev to find out which dump device is actually in use. You can also use this command to find out how big the dump device has to be ( -e , estimate) and to set the dump device ( -Pp <device> ).

On another note, you seem to have doctored with the rootvg because it is quite unusual to have a 1GB PP-size. Usual PP-sizes in rootvgs are indeed 64MB and 128MB. I don't know what exactly you did, but: might it be that this has something to do with it?

I hope this helps.

bakunin

1 Like

I figured out the problem. Yes I had already used sysdumpdev to verify that I was actually using the devices in question. The rootvg is fine, the reason it has a 1GB PP size is because the disks it was installed on are 4TB disks. So I believe it defaulted to 1GB PP size.

The problem was with the dumpcheck script. I ran it with debug on and found that it was reporting the block size for the dump device as 512 bytes. When I knew in fact that the block size was 4k. I found that it was an old version from 2010 that was missing a lot that the version on on most of our other servers were using. I copied the later version of the script, which correctly specified the block size of the dump device, and voila. So really, the dump device was fine, just the dumpcheck script was reporting the wrong size.

Thanks for your reply!

1 Like

OK, this is a valid explanation. Still, i suggest you use smaller disks for your rootvg. In my experience it is best to put only filesystems really really belonging to the system (not the application, not the data, not anything else) into the rootvg. First, when you take a system backup you back up the rootvg, aka mksysb . Guess, how long that takes on a multi-terabyte rootvg and, bonus question, how big of a size the resulting mksysb-image will be. For reference, my largest database server (700G memory, ~80TB database) has a (2x, mirrored) 120GB rootvg - and this is only because the customers insisted on swap space that will never be used. Otherwise i would have gone with our 2x60GB-standard-rootvg.

Second, modern systems are usually virtual and the disks are too. There is a big difference in moving around system disks (controlled by VIOS) and non-system disks because these can be unmounted/ varyoff ed easily while the system is under load. Therefore it is a good idea to separate system- and non-system-disks into seaparate VGs.

First off: thank you for telling us that! I perhaps never would have had that idea at all and so i (and all the others reading this thread) learned something from here too. Absolutely commendable! A word of caution too: you shouldn't just overwrite parts of the system. AIX has a great packaging system and i suggest you use it to your advantage:

  • check which package the file came from with
lslpp -w </path/to/file>

then replace this package with a newer version.

  • if you are interested in which else files the package contains issue
lslpp -f <packagename>

to get a list of files/directories belonging to this package.

I hope this helps.

bakunin

1 Like

Thanks! Yes, I agree this system should not have been installed on such large disks. There are a couple of 300GB disks in there that I think the rootvg should go on. I am new to this company and wasn't involved in installing the makesysb image. However, it is new and not in production yet, I may make the recommendation we move it to the other disks. Most of our systems do use virtual disk, however I think due to the particular function of this server they decided to install locally.

Thanks for the info about the packages, I will look into that! I have been a Unix/Linux SA for 15 years, but am very new to AIX, so I appreciate the advice.

------ Post updated at 10:06 AM ------

Just for some additional information, it looks like this problem existed on all the servers we had running AIX 7100-03-04-1441. The fileset for the dumpcheck script was bos.sysmgt.serv_aid 7.1.2.0.