UNSOLVED

dmitri-cml

updated

14 years ago

DC

dmitri-cml

1 Message

0

14488

August 17th, 2012 05:00

Some beginner raid type/volume behavior questions

- Why is a volume formatted with VMFS with all VMs moved out of it, still shows space used?

 - If volume is re-assigned from raid10 active/raid5/6 replay profile to raid5/6 only profile, when should we expect the volume to start releasing disk space?

 - Does it need data replay activated for both above scenarios?

 - Why Compellent shows more space used on all VMFS volumes, compared to what vcenter client shows? Ex: vol1 (1,143.85Gb used per vcenter, 1,730Gb used per CML Storage Center)?

 - All volumes on Tier3 are currently using raid10-dm/raid6-10 scheme. Is it possible to convert some volumes to raid10/raid5-5? Can I specifically create a custom storage profile with say raid10 (non-DM) active and raid5-5 for replay? The only choices are raid10 and raid 5/6 for replay, raid 10 only, and raid5/6 only.

Thanks in advance! 

  • hallidayr

    48 Posts

    1300

    0

    Posted August 17th, 2012 10:00

    Welcome to Compellent! Sounds like you could benefit greatly from their training course. I went through this a couple months after we got the systems and it helped a lot. There's a bit of background knowledge but I'll try to explain the best I can.

    1. The reason for this is how the Compellent SAN virtualizes its storage. The design is such that no space is actually used on the disks until data is written there. VMFS is just like any other file system in that data is not removed when files are deleted. The entry of their existence is simply removed from the filesystem's database, journal, file allocation table, or whatever you'd like to call it; the data is still there. vSphere 5 introduces support for a SCSI_UNMAP command which will tell the SAN to actually free the blocks when a file is deleted. Unfortunately this is only in Storage Center 6 in Compellent world, so until this gets released for the older controllers it's only available on Series 40.

    2/3. You need Replays and Data progression set up for this to work. It will begin to restripe next cycle (default is 7 PM I believe) but you may see "allocated" blocks for a while. I've had to call Copilot a few times on issues like these.

    4. This comes back to point #1 as well as RAID overhead. The exact space usage is highly dependent on your RAID, block size, and replay profiles. For example I have a 300 GB mail store that's using 1.5 TB worth of disk. Why? Every night there's about 90-120 GB of churn captured in replays. The most recent data is kept at RAID10 (so x2), replays are at RAID 5-5 (more space there), and on top of it the filesystem just puts files wherever it wants so the SAN sees a lot of used blocks. In addition to this, with a 2 MB block size, a 5 KB email coming in will have an on-disk footprint over 400 times its size before RAID. With VMFS similar churn can come from VMware Snapshots, powering on/off VMs (swap file creation), Storage vMotion (don't do this on Compellent!!!), or migrating VMs between data stores.

    5. Perhaps - a lot depends on your hardware configuration. Refer to the manuals on how to prepare a new storage type, this is also where you set the block size.

  • 1300

    0

    Posted August 23rd, 2012 17:00

    In response to your last question, you're looking for the capabilities introduced in manual storage mode. As hallidayr pointed out, you should definitely attend their training course.