mminfo is ... confused?

[benzino1]

have a look at this:
mminfo -a -c client2 -c client1 -q'sscreate>="07/01/09 12:00:00",sscreate<="07/31/09 23:59:59"' -r'name,client,totalsize(1)'| wc -l
15618
mminfo -a -c client1 -q'sscreate>="07/01/09 12:00:00",sscreate<="07/31/09 23:59:59"' -r'name,client,totalsize(1)'| wc -l
878
mminfo -a -c client2 -q'sscreate>="07/01/09 12:00:00",sscreate<="07/31/09 23:59:59"' -r'name,client,totalsize(1)'| wc -l
2365

Am I so dumb that I cannot add to simple values (2365+878!= 15618) ? Or there is something wrong with mminfo?

[DavidHampson]

No, you are dumb for a different reason - take a look at the client specified

[DavidHampson]

It looks like I am the dumb one and didn't read your mminfo correctly Also you have the following field reported on:

totalsize(1)

This will return the first character of the totalsize field. Each saveset instance will be reported the same for name and client and have a 1 in 10 chance of being the same value of totalsize(1) so when you do your wc command there is a chance some of these instances will be the same. If your clients back up to different servers then there is a possibility there will be some duplication in these lists but then we would expect both clients reported on to be either the same or lower than the combined value...

I'm going to test it on a couple of clients here and see what happens...

What exactly is it you are trying to report?

[benzino1]

totalsize(1) will return me info with the information whether it is KB,MB,GB not the first character of the totalsize:
client1 /path_to_something1 254 MB
client1 /path_to_something2 14 GB
client1 /path_to_something3 987 KB

To tell the truth all I want is a report, just as you can see above. But it is really silly that this report differ when I'm asking mminfo for report for clients separately and the report for both of them together.

[DavidHampson]

I expected the totalsize field to be truncated... I ran a query with two randomly chosen clients and got the same results as you did - 3259+1292 = 9564 and I got the same result with totalsize and totalsize(1)

[benzino1]

Don't you think it's ... strange (at least, not to say BUG)?

[DavidHampson]

Very strange - I've taken a look at some of the output but the strange thing is there is discrepancies on both sides of the output. Since we would expect the output to be consistent then I would call this a BUG...

[ark-flUcr]

hello;

We have similar problem. EMC souport commented that this could be related with:

LGTsc29876 "incorrect mminfo output when client is part of the query"

They are currently investigating.

PROBLEM DESCRIPTION:
------------------------------------
hello;

when we use nsrinfo to find a file, we see the date is the same as the mminfo query.


# mminfo -q" client=CLIENT,savetime >= -60 day,level=full" -r nsavetime,savetime,name,level |grep " / "
1244283022 06/06/09 / full
1246698992 04/07/09 / full
# nsrinfo -N '/etc/hosts' CLIENT
scanning client `CLIENT' for all savetimes from the backup namespace
/etc/hosts, date=1246698992 sáb 04 jul 2009 11:16:32 CEST
/etc/hosts, date=1244283022 sáb 06 jun 2009 12:10:22 CEST
...
...
/etc/hosts, date=1217669200 sáb 02 ago 2008 11:26:40 CEST
/etc/hosts, date=1215237853 sáb 05 jul 2008 08:04:13 CEST
22 objects found
#

But we have a problem:

# mminfo -q" client=CLIENT,savetime >= -60 day,level=full" -r nsavetime |grep 1244283022
1244283022
# mminfo -q" client=CLIENT,savetime >= -60 day" -r nsavetime |grep 1244283022
#
why in the second query (ite would have to give more information) we do not see all information???

And if we do these querys; we see the query for 160 days have less information than the query for 30 days:
# mminfo -q "savetime >= -30 day,client=CLIENT" -r "name,nsavetime,savetime,level" -o t|wc -l
1436
# mminfo -q "savetime >= -60 day,client=CLIENT" -r "name,nsavetime,savetime,level" -o t|wc -l
1491
# mminfo -q "savetime >= -160 day,client=CLIENT" -r "name,nsavetime,savetime,level" -o t|wc -l
1220
#

[DavidHampson]

I ran tests both with -c client and client=clientname and got the same results. So to be more accurate until this resolves it looks like report on all clients and grep the results...

[benzino1]

No, we cannot, and I tell you why. We cannot trust the results of mminfo anymore. There is no assurance that there is no any other problem. We find out that there is a problem when client is specified on the command line, but now we are not sure whether there is no problem with other switches. Let's say that someone wrote a script to do something odd (delete volumes, label volumes, expire savesets or do anything else which cannot be undone) based on the mminfo output - Ooops sorry but there was a bug and you've lost your backup which is critical for you, "have a nice day and see you later"