Subscribe

question regarding snapshots

hi

i have implemented snapshots on a fairly large volume.

The volume size is around 2.5 TB.

It has a lun of around 800 gb.

The snapshot size is fairly reasonable around 25 GB , however i noticed that when there is a gap of a couple of days, the snapshot size jumps to aroound 500 gb which is like 20 times the normal snapshot size if taken daily.

Why is this so?

Is there anything that i can do to avoid such large size snapshots.

What would be the best recommendations to lower the size of the snapshots.

Moreover in order to avoid the volume going out of space , i have set a fractional reserve of 100% and also set dedeuplication on it .

Is there anything more i can do?

Re: question regarding snapshots

Hi Rakesh - Your snapshots are more than likely larger after a few days, due to the data change rate for the LUN, as all the new data blocks are written to it. A snapshot size is determined by the amount of block changes that have occurred between 2 snapshots. If your daily snapshot is approx 25GB then this is the amount of data block changes during this period.

What data is held within LUN? To stop the volume running out of space, you could set the volume autogrow options, however at 2.5TB you may want to keep an eye on the space.

have a look at this YouTube video to get an idea of how snapshots work:

http://www.youtube.com/watch?v=JHfGKN0b9gU

Martin

question regarding snapshots

Hi Martin,

That is fine , i have already set the volume to autogrow.However setting the fractional reserve to 100% , wouldn't that guarantee me that my lun is never taken offline for the lack of space in the volume.

My lun is basically a lun given to a xen server which as a filesystem for lotus notes.

Re: question regarding snapshots

Fractional Reserve would protect your LUN, as it provides over right space, for any large data changes rates. However it would never fully prevent the LUN going offline. If the containing aggregate was to reach the max capacity, the aggr would be taken offline to preserve and protect the data within it.

Martin