aggregate snap reserve size

[slava]

Hello guys,

I would like to know what is the minimal aggregate snap reserve size i may left on the system?

[tyeoman_swinton]

Hi,

I've reduced this to 2% before, but be very careful with reducing aggr snap reserve as it gets used for a few different things.  One example is semi-sync SnapMirror where it is used as a buffer.

[slava]

Thank you Yeoman,

In my case there are no any snapmirror at all.

Slava Chubar

Professional Services Engineer

NetApp

972.3.9205538 Direct

972.52.3120538 Mobile

slava@netapp.com

www.netapp.com

[pascalduk]

If you don't have RAID syncmirror and / or a metrocluster configuration and you don't expect to use aggregate snap restores (has anyone ever used this???), you don't need snapshots on aggregate level and therefore can set the aggregate snap reserve to 0%. If you use aggregate snapshots set the snap reserve to such a level that you normally don't run of reserve space.

We are running on all our non metrocluster configurations with 0% aggregate snap reserve and no aggregate snapshots (snap sched -A <aggregate> 0) ever since we migrated to ontap 7.0.

[slava]

Thank you Dukers,

As I know, you need to save some space for some maintenance processes like block reallocation or in case of aggr inconsistence when you need to run wafl_check.

Slava Chubar

Professional Services Engineer

NetApp

972.3.9205538 Direct

972.52.3120538 Mobile

slava@netapp.com

www.netapp.com

[amiller_1 · Accepted solution]

I wouldn't go below 3% aggregate snap reserve (that does let you get back 2% from the default at least) as doing otherwise can cause data corruption issues (as well as cause issues upgrading to 7.3.x).

See this bug report as well....

http://now.netapp.com/NOW/cgi-bin/bol?Type=Detail&Display=164329

There was also a general warning in the 7.3.x upgrade docs about this as well giving 3% as a minimum.

I would stay away from 0% aggregate snap reserve as I have personally had data corruption issues with this (a volume with "file" space guarantee got set to "none" silently and subsequently had some corrupted files which we didn't notice for a couple months due to some odd circumstances).

[pascalduk]

The bug is not about snap reserve size for the aggregate, but about free space in the aggregate. Using a snap reserve of more than 0% is only a tool to make sure nothing can ever take up all the space in an aggregate.

I always have at least 10% free in the aggregate, therefore for me this will never be an issue. But if I am wrong I would really like to know about it.

[evilensky]

This thread is a top Google result for "aggregate snap reserve"

  Have you come across any contrary information?

[pascalduk]

There is no right or wrong here, just do what makes you comfortable.

If you don't use aggregate snapshots, there is no need to to reserve space in the aggregate. But some use the snap reserve as a tool to make sure there is always (enough) free space in the aggregate.

[aborzenkov]

http://now.netapp.com/NOW/cgi-bin/bol?Type=Detail&Display=164329

I must honestly admit I fail to understand what exactly was fixed. Bug description visible to me sounds like deliberetae design decision. Do you have any details (i.e. what exactly was changed in a "fix")?

[jayadratha]

Hi, we always leave default 5%.

And disabling aggr snap reserve only with metrocluster configuration or when extra space needed (not best practice but...).

http://now.netapp.com/NOW/knowledge/docs/ontap/rel727_vs/html/ontap/sysadmin/provisioning/concept/c_oc_prov_sys-maintenance-aggr-snapshot-copy-managin...

Best practice for us is at least 10% free space in aggregate.

[mheimberg]

Hi

jayadratha schrieb:
And disabling aggr snap reserve only with metrocluster configuration

This is definitely wrong: The MetroCluster uses SyncMirror and this in turn uses aggregate snapshots:

"...However, SyncMirror aggregates use snapshots to optimize plex resynchronization...."

MetroCluster is the one configuration where one should absolutely *not* disable aggr reserve (set to 0%).

Markus

[jayadratha]

Hi!

I need to say "disable automatic aggr snapshot creation by scheduler".

Because sycmirror don't use this aggr snapshot copy.

[mheimberg]

Hi

I need to say "disable automatic aggr snapshot creation by scheduler".
Because sycmirror don't use this aggr snapshot copy.

Yes, this way I completely agree

So it will be:

snap reserve -A aggr0 5

snap sched -A aggr 0 0 0

or

aggr options aggr0 nosnap on

regs

Markus

[mikesicarup]

I asked a NetApp instructor about this, and besides the Synch mirror, the aggr snapshots are leveraged by WAFL check and WAFL iron.  He indicated that if you need to run these utilities, not having aggr snapshots would take days as opposed to hours.  It made me take pause.

[DRUMDUDESAN]

Hi,

Can we get an update on from NetApp themselves on this please running NetApp Release 8.1x and above.

What is aggregate snap reserve required for?

What are the implications of setting snap reserve on an aggregate to 0% that does not hold DataOntap/vol0?

Thanks

Jeff

[adaikkap]

Hi Jeff,

     vol0 is on an  aggr and not on the snapreserve of an aggr.

Regards

adai

[DRUMDUDESAN]

Hi Adai,

You have not understood the question and the post thread.

[thomas_glodde]

Aggregate snapshot reserve is only needed in metrocluster configurations, its set to 5% then.

in all other configurations you can drop it to 0%, its even the new default for ontap 8.1 and beyond.

[DRUMDUDESAN]

Hi,

Where is this located in the documentation or recent TR document?

Thanks for your response.

Jeff

[thomas_glodde]

https://library.netapp.com/ecmdocs/ECMP1120690/html/frameset.html

ontap 8.1 documentation

Root aggregate Snapshot reserve values
reduced

The root aggregate and root volume Snapshot reserve values
for new storage systems have been reduced starting with Data ONTAP 8.0.1.

To increase the amount of usable space on the storage system while
still retaining all the benefits of volume Snapshot copies, NetApp ships all new
systems running Data ONTAP 8.0.1 or later with the following new Snapshot
reserve values:

• Snapshot reserve for the root aggregate is set to 0%.
  The previous value was 5%.
• Snapshot reserve for the root volume is set to 5%.
  The previous value was 20%.

Existing systems, and systems configured at the factory to run
MetroCluster, are not affected by this change.

Attention: If you later
configure MetroCluster, or use RAID SyncMirror or the aggregate copy capability
on the root aggregate, the root aggregate Snapshot reserve must be increased to 5% by using the snap reserve command.

For more information about using the snap reserve command, see the na_snap(1) man
page.