This post is more than 5 years old
68 Posts
0
1682
adv_file did not recover space
Hello,
I've got an adv_file (about 10 TB). After having deleted yestarday all the savesets on this adv_file (with nsrmm -d -S), I cannot see that my volume is empty.
I've tried nsrim -X (even if an automatic runs on the night) and nsrstage -C -V, so I don't know how to get space again (this sunday, I'll need about 8 TB to make big saves...).
Here are my results (NW 8.1.1.3 under french AIX 7.1) :
# df -g /adv01
Système de fichiers Blocs GB Libre %Util Iutil %Iutil Monté sur
/dev/adv01lv 9995,00 5433,10 46% 207 1% /adv01
# mminfo -avot -r "client,name,totalsize,nfiles,savetime(20),ssflags" PB152101=DADV01.001
6095:mminfo: no matches found for the query
# mminfo -avot -r "written" PB152101=DADV01.001
4779 GB
I do not want to relabel the volume because I'll lose all references to cloned savesets sourced from this volume.
Any clue/solution ?
Thanks in advance.
Denis
ble1
2 Intern
2 Intern
•
14.3K Posts
0
June 3rd, 2015 04:00
Having = in name is asking for trouble.
Anyway, if we assume that mminfo which doesn't show ssid is correct (forget written part as that was never really accurate), then when you to disk device and you do ls -laR you will see name of ssids in long format name. From there, you can query it via mminfo and see what they are exactly and what is their status.
bingo.1
2.4K Posts
0
June 3rd, 2015 02:00
- "written" is also a statistical parameter. Just because you have deleted a save set does not mean that this info will change right away.
- The physical deletion of save sets could have a delay of one day, depending on the start of the appropriate processes.
- NW treats the volumes individually. Just because you label a media should not affect other instances of the same save set.
- Obviously, PB152101=DADV01.001 is your volume name. It seems to be valid. However, let me recommend that you only use characters/separators which are listed in the label templates. Others could be misinterpreted.
- Let me suggest that you use "nsrim", not "nsrim -X". See example below.
########### NW 8.2.1.1 (Windows) ############
D:\>nsrmm -d -S 4285447855
Delete file and media index entries for save set `4285447855'? y
D:\>mminfo -avot -r "written"
5000 MB
D:\>mminfo -avot
6095:mminfo: no matches found for the query
D:\>nsrim
86069:nsrim: Processing 1 clients
86068:nsrim: Managing 1 volumes.
AFTD_1: 5000 MB used, 0 save sets, appendable
86073:nsrim: Compressing media database.
D:\>mminfo -avot -r "written"
0 KB
D:\>
deniscnqd
68 Posts
0
June 3rd, 2015 04:00
Having = in name is asking for trouble.
Used for years without problems ;-)
Thank you for the exploration of the filesystem, I've found some ssid but they seems fine...
#mminfo -avot -q "ssid=b8ee9670-00000006-7559aa6d-5559aa6d-a28c2604-9a6ff402" -r "client,name,ssid,cloneid,totalsize(5),savetime,ssflags,volume"
client name ssid clone id total date ssflags volume
client1 /dbs2/ISMM42ZE 1968810605 1431939693 40 GB 18.05.15 vF PB152101=DADV01.001
I'll try to delete some and see if I can get space.
deniscnqd
68 Posts
0
June 3rd, 2015 04:00
As I said, deletions happens yesterday (and for some, before yesterday).
nsrim (with no parameters) shows me :
But volume seems to be still populated :
# mminfo -avot -r "written" PB152101=DADV01.001
4779 GB
NW treats the volumes individually. Just because you label a media should not affect other instances of the same save set.
First times I relabeled the volume, I lost my tape clones in the media database
ble1
2 Intern
2 Intern
•
14.3K Posts
0
June 3rd, 2015 05:00
Some people have sex for years without problem too and then it happens
Your mminfo query doesn't show until when this is supposed to be kept, but bare in mind that it is backup made on 18th so if your retention is 1 month it should be there a bit more. Also, you didn't include level information and you do not wish to break your backup cycle by removing something that is needed for full restore when doing PiT.
deniscnqd
68 Posts
0
June 3rd, 2015 06:00
Some people have sex for years without problem too and then it happens
...or not
So I've took all long-ssid under my filesystem then deleted them.
After a nsrstage -C -V, I've got my space back...
Got some empty directories 'notes' under my filesystem but the whole space is available (9991,47 GB on 9995 GB)
Big thank !
I use adv_file to speed up saves (concurrent and not using tapes). After the saves have finished, I clone on tape (with a long retention) then delete the ssid on the adv_file, that's why retentions on the adv_file are short.
The question that remains is : why theses ssid where not shown when querying with mminfo ?
ble1
2 Intern
2 Intern
•
14.3K Posts
0
June 3rd, 2015 07:00
For why part you need debugging. One question, why clone to tape and not stage given that you remove it after that? Retention you can set on pool so when moved to tape pool it will get higher retention (ok, not sure if that works with staging, but with backup it does so I would assume same applies to staging). I used to have similar setup in 2003 and I would usually stage via script by doing nsrstage (nowadays nsrclone -m) per device. Back in those days you could not set retention on pool so I simply used same ssid list I previously generated and run something like:
for ssid in `cat ssid.lst`;do nsrmm -y -e '+1 month' -w '+1 month' -S $ssid;done
deniscnqd
68 Posts
0
June 3rd, 2015 07:00
For debugging, I'll try the next century...
I don't stage because since clones tapes are externalized (for security), I want to keep saves online between two save sessions (2 times a month) for speed up recovery if needed.
I manually delete savesets the day before the new save.
ble1
2 Intern
2 Intern
•
14.3K Posts
0
June 3rd, 2015 08:00
Ah I see... I guess you have single tape/location... we had 2 so I would stage, reset policies and then clone from one PTL to another and if I had to export anything (which I had at the time) then secondary copy would be taken out.