Other than being used as tier in FAST VP, are they also being used for provisioning?

When used with FASTVP, you only need to bind to one tier/pool. In your case you are binding TDEVs to the FC pool, and then you apply a FASTVP policy to the devices. The Policy defines the rules under which FASTVP can move portions of the TDEVs into the other tiers.

You will end up with the FC pool appearing to be oversubscribed, but having free space, and the EFD and SATA tiers having 0% subscription, but having lots of data in them. This is normal for FASTVP and binding all TDEVs to the FC tier is recommended.

Thank you both for sharing.

Is this commonly practice to use SATA as just one of tiers in FAST VP?

How do I verify that SATA devices are not being used for provisioning at all. If I understand correctly, people could use them as STD to provision them as regular way.

emcmagic wrote: Thank you both for sharing. Is this commonly practice to use SATA as just one of tiers in FAST VP? How do I verify that SATA devices are not being used for provisioning at all. If I understand correctly, people could use them as STD to provision them as regular way.

remember that STD devices are built on top of disk groups, not pools

that's correct, Quincy56.

With FAST VP, the VP policy will take care of which of the 3 pools (or tiers contained in the policy) to get the thin device extent from and you'd only need to bind the TDEV to 1 of the available pools, generally the FC pool/tier.

Like you said, if you know that the nature of the workload is mostly idle, then creating TDEVs and binding them to the SATA pool would suffice.

Thanks.

Hello emcmagic,

If you've created TDEVs and have bound them to the FC pool/tier, you could still have the TDEV using extents from the SATA pool/tier. This is when the FAST VP controller thinks you have extents that can be moved to the SATA tier because they're mostly idle.

To verify if there's a TDEV that is using the SATA (or the EFD tier) :

# symcfg show -pool <> -thin -detail -gb -sid XXX

Scroll down and Look at the "Other pool bound thin devices" section for the SATA pool .. If it lists one or more TDEVs, these are the TDEVs that have actually come from another pool/tier (the FC tier in this case) and are using extents from the SATA data devices (TDATs)

[In short, a TDEV bound to the FC tier is using extents from the SATA tier]

Another way to check if the TDEV (that is bound to the FC pool) is asking for extents from the SATA pool :

# symcfg list -tdev -range 5EcB:5EcB -bound -detail -v

Thanks.

just for your reference ..

Another way to check if the TDEV (that is bound to the FC pool) is asking for extents from the SATA pool (and/or the EFD pool):

# symcfg list -tdev -range 016F:016F -bound -detail -v

This is how, the output would look like, for TDEV 016F, that is bound to 1 pool/tier, but is requesting tracks from the other 2 pools contained in the policy.

Thanks.

Hi rohang,

Your messages are very helpful, and beyond what I am questioning on. Appreciate your help.

To make sure I understand you correctly. In your case, TDEV is bound to SATA_FASTVP, and asking, and asking only (not allocated yet) for extents from FLASH_FASTVP, and FC_FASTVP. correct?

the following is the result of running "symcfg list -tdev -range 0fa3:0faf -bound -detail -v". In my case here, TDEV is bound to FC_T1, and extents are actually allocated in ATA_T2. correct?

                  S Y M M E T R I X   T H I N   D E V I C E S

-------------------------------------------------------------------------------

                                   Pool        Pool          Total

                  Flags     Total  Subs     Allocated       Written

Sym  Pool Name    EMSP     Tracks    (%)  Tracks  (%)  Tracks   (%) Status

---- ------------ ----- --------- ----- --------- --- --------- --- -----------

0FA3 FC_T1        FX..    2209680     0         0   0   1269492  57 Bound

     ATA_T2       --.-          -     -   1269624  57         -   - -

0FA7 FC_T1        FX..    2209680     0         0   0   2207778 100 Bound

     ATA_T2       --.-          -     -   2207844 100         -   - -

Glad I can help, emcmagic.

please mark the answers as helpful

Yes, In your case, the TDEV 0FA3 is bound to FC_T1 and the extents for this TDEV are being allocated from the ATA_T2 (the SATA tier). The written % for FC_T1 is 57 and the allocated % for ATA_T2 is 57%, which means all the extents that this TDEV is using, are allocated from the SATA tier.

Similarly, for the second TDEV 0FA7, all extent allocations are coming in from the SATA tier. (And the TDEV is full - i.e no more extent allocations allowed). Also, the storage group that this TDEV is a part of, is associated to a policy that must have specified a maximum of 100% allocation (the default, if not specified) from any of the tiers contained in the policy.

[something like a .. #symfast -sid XXX -fp - add -tier_name ATA_T2 -max_sg_percent 100]

Again, this is a changeable parameter. You may modify how much of what tier, a TDEV may use .. using a #symfast -sid XXX  -fp_name -modify -tier_name   -max_sg_percent

In my case, the TDEV is asking for, and is being allocated tracks from the other 2 pools. The TDEV is bound to the SATA tier but is being allocated 7% of (the 59%) tracks from tiers other than the SATA tier.

Remember, the "written tracks" value is reported against the pool that the device is bound to, whereas the "allocated tracks" value is reported against all the pools (inside the policy) that the TDEV is asking for tracks from.

Thanks ..

Hi rohang, I appreciate so much your time, and you are it more steps further to address all my questions in my mind. I marked all your messages as helpful. Thank you!

Sorry, allow me to mark only one.

Hi rohang,

I am sorry, but one question was left in my mind about your comment below:

>>And the TDEV is full - i.e no more extent allocations allowed

Does that mean the lun is associating with this TDEV is getting full, and if all DATA devices that TDEV points to are full, then I assume I can not assign new DATA dev, in this situation if writes continue, then the will host be getting error? I read somewhere else saying TDEV full should be okay.

I know TDEV is a cache device, and contains all pointers, is there any way to increase the size of it?

Hello emcmagic.

If a TDEV shows up as 100% under a "pool allocated tracks", it means that the TDEV has fetched as much extents (in increments of 12 tracks or 768 KB) as it could, from all available Datadevs that reside in the pool. This would not necessarily mean that the host that is using this TDEV would receive an error when it would try writing to this LUN. There could still be some free space on the TDEV because the application or the user is deleting files and the filesystem is re-using the space, which would otherwise result in a considerable amount of fragmented space on the LUN.

When the user creates a new file on the host filesystem, depending on the filesystem, it may or may not use the space freed up by a deleted file. If it writes the new file in the previously freed-up area, then the thin LUN will not consume more space from the pool. However, if it writes in a previously unwritten area, then the thin LUN will consume more space from the pool (by requesting extents from the Datadevs) in which case, the host will receive an error.

NTFS is not very effective when it comes to reusing deleted space. Other file systems like Linux ext2 and ext3, Solaris ZFS and UFS, and Symantec VxFS effectively reuse deleted space before allocating new space and thus work nicely with thin LUNs. For VMware environments, a VMFS datastore reuses previously allocated blocks that are beneficial to thin LUNs.

To increase the size of a TDEV, you need to convert it into a meta.

# symconfigure -sid XXXX -cmd "form meta from dev 5ECB, config=concatenated; add dev 5ECC to meta 5ECB;" commit -v

(where 5ECB is your original TDEV)

thanks ..

Hi rohang,

Thank you very much for sharing. It helps a lot!