Subscribe
Accepted Solution

Aggregate Almost Overcommitted

We are getting this message in Operations Manager for one of our lone aggregates. It's not part of resource pool. Thresholds are set to defaults for the aggregate and for the global settings (so Aggregate Nearly Overcommited Threshold is 95%). Snapshots are currently disabled. Double-Parity is checked. I'm not sure if Thin Provisioning is used or not. The condition says:

"Committed 7.32 TB (99.60%) out of 7.35 TB available; Using 5.84 TB (79.44%) out of 7.35 TB available"

The "Space Breakout" is thus:

Aggregate Size7.35 TB
Snapshot Reserve Size837 GB
Used Space5.84 TB
Available Space1.51 TB
Available Snapshot Reserve526 GB
Used Snapshot Space311 GB
Unused Guaranteed Space2.41 TB
 

We added two additional disks to the aggregate yesterday to bring the used capacity percentage down from 84% to 79.44%. I guess it didn't help the Committed percentage though, which must be unrelated.

So what exactly does 7.32TB commited mean? How is that calculated? How can we reduce that value to get it back under 95%? Let me know if there are some command-line commands that can be more enlightening than the GUI. How can we tell if thin-provisioning is in use?

Sorry if this is in the wrong forum (please help move along if necessary), first time poster.

Thanks!

Re: Aggregate Almost Overcommitted

Hi

Welcome to the forums

First performance will suffer once the aggregate gets above 85% full and the higher this gets the worse it will become.  This is because the filer is hunting for free block to write too and is constantly reclaiming them.

I notice you have an aggregate snap reserve set of 5%.  Many people turn this to zero as for them it is wasted space.  We would not aggregate snap restore, only at the volume level.

aggr snap reserve 0

This would give you 837 Gb back

Yes you are thin provisioned.  Have a read of this for what is going on.  http://blogs.netapp.com/shadeofblue/2008/10/really-thin-pro.html

Also be careful adding in disks one at a time as you will create hot spots in your aggregates with all the IO going to the new disk until it is as full as the other disks.

Hope this helps

Bren

Re: Aggregate Almost Overcommitted

Hi Brendon,

Thanks for the info on performance problems over 85% usage, snap reserve, thin provisioning, and adding disks (we actually added two more at the same time, not sure if that created a two disk hotspot or not). That does help a little.

I guess I'm still confused on aggregate overcommittment and how we reduce this value to get back under the threshold? I gave it more disks, but that did nothing for committment. Removing the snap reserve also seems to be just creating more "free space" and helping us stay under the 85% usage. Unless I missed something, we haven't addressed overcommitted yet, have we?

Thanks!

Re: Aggregate Almost Overcommitted

On the CLI, use "aggr show_space" to show true space allocation/usage for the aggregate in question. That output will also show you if space guarantees are in use.

Sean.

Re: Aggregate Almost Overcommitted

Hi Michael,

     Can you get the output of the following report for your aggr ? To see why it is overcommitted ?

dfm report view volumes-space-guarantees <aggr name/id>

Regards

adai

Re: Aggregate Almost Overcommitted

> aggr show_space

Aggregate 'AGGR01F_VMDATA_H03'

    Total space    WAFL reserve    Snap reserve    Usable space       BSR NVLOG
   9747745280KB     974774528KB     877297072KB    7895673680KB             0KB

Space allocated to volumes in the aggregate

Volume                          Allocated            Used       Guarantee
VM_CONFIG_FILES_H03            78955744KB      55719244KB          volume
VM_SNAPSHOT_H03               131141424KB         79452KB          volume
VM_SWAP_H03                   132449216KB      12936640KB          volume
VM_T1_PRI01_H03               504132756KB     329457160KB          volume
VM_T1_SEC01_H03               928648876KB     374106984KB          volume
VM_T1_OSSWAP01_H03            185736048KB     138311856KB          volume
VM_T3_OSSWAP01_H03            132669036KB      68442612KB          volume
VM_T3_PRI01_H03               822486052KB     720505312KB          volume
VM_T3_SEC01_H03              1034757948KB     898116724KB          volume
VM_TEMPLATE_H03               524875800KB     231609016KB          volume
VM_T3_SEC02_H03               918863184KB     340049036KB          volume
VM_T3_SEC03_H03               918802796KB     565447436KB          volume

Aggregate                       Allocated            Used           Avail
Total space                  6313518880KB    3734781472KB    1623106244KB
Snap reserve                  877297072KB     348884024KB     528413048KB
WAFL reserve                  974774528KB        593540KB     974180988KB

> dfm report view volumes-space-guarantees AGGR01F_VMDATA_H03
dfm not found.  Type '?' for a list of commands

Re: Aggregate Almost Overcommitted

you must qualify the aggr name with the filer name as filername:aggrname or better use the id of the aggr which can be obtained from

dfm aggr list

Can you also get the report for the following

dfm report view volumes-autosize-details < aggr name/id> as autogrow turned on is also considered for aggr overcommmit

calculation.

Re: Aggregate Almost Overcommitted

I don't think we have the dfm command. I'm puttied into the filer on the command line. Is dfm a (client) product that should be running on my windows pc or a linux box?

> ?

<snip>

cifs                igroup              ping6               stats
config              ipsec               pktt                storage
date                ipspace             portset             sysconfig
df                  iscsi               priority            sysstat
disk                keymgr              priv                timezone
disk_fw_update      license             qtree               traceroute
dns                 lock                quota               traceroute6
<snip>

Is there another command-line command to show the auto-grow feature per aggregate? I'm pretty sure we used this at one point. Not sure if this was at the volume level like dedupe, or the aggregate level though.

Thanks!

Re: Aggregate Almost Overcommitted

Yes, DFM is a separate product.

Sean.

Re: Aggregate Almost Overcommitted

yes dfm can run on windows or linux.

autogrow is an options at volume level.

You will have to drill down to each volume of the aggr.Its not available at aggr level.

Regards

adai