About aborzenkov

aborzenkov · ‎2010-12-20

Fragmentation? What is the result of "reallocate measure" for volume on secondary system? You could run it periodically for some time to see trend. It would also be interesting to run (at least once) "reallocate measure" for each individual LUN, as some of them could be more fragmented than others.

aborzenkov · ‎2010-12-20

It is possible only if your switches support link aggregation (stackable switches, VSS, vPC ot similar). Otherwise you are limited to single mode VIF which is active/passive.

aborzenkov · ‎2010-12-20

This is actually DataONTAP issue and was introduced somewhere around 7.2.3. DataONTAP would force fractional reserve to 100% unless guarantee is set to volume. This should be fixed as of 7.3.3: "You can only set fractional reserve for a volume with a space guarantee of none with Data ONTAP version 7.3.3 and later and version 8.0.1 and later."

aborzenkov · ‎2010-12-20

Does it happen for this particular qtree only or for any qtree with NTFS security? NetApp always performs NT-to-Unix user mapping, even for access to NTFS qtree from Windows client. If mapping fails, access is denied. Check, that - usermap.cfg does not deny access by listing empty Unix user name, like \ => "" Any NT user which maps to empty Unix user in this way will be denied access - you have non empty wafl.default_unix_user. Default is pcuser that is normally available in /etc/passwd

aborzenkov · ‎2010-12-18

What exactly do you mean under "user needs to appear in /etc/passwd"? What does not work if user is not entered there?

aborzenkov · ‎2010-12-16

There is a corner case – while free blocks are not transferred, freeing blocks is not instantaneous. For a large file it may take significant time. So if SM was initiated while free block scan is not yet complete, blocks not yet marked as free may be transferred.

aborzenkov · ‎2010-12-16

Windows user is mapped to Unix user and then file access is checked using standard Unix rules. That also answers your second question about groups – at the time permissions are checked NetApp is not even aware that was a Windows user. Group membership is determined using /etc/group, NIS or LDAP, whatever is configured. You can check tr3490 for overview of multiprotocol access rules.

aborzenkov · ‎2010-12-16

AFAIK connecting shelves via switch is not supported, irrespectively of port mode. The only case when shelves are connected to switches is MetroCluster, which is restricted controlled environment. It may appear to work, but I would really speak with NetApp representative regarding future support of such configuration.

aborzenkov · ‎2010-12-11

This applies to V-Series LUN, not to native disks. You could also look at NetApp kb34044 (although I am afraid they changed document numbers recently).

aborzenkov · ‎2010-12-11

There are also aggregate level snapshots.

aborzenkov · ‎2010-12-11

a) Aggregate size is shown base 1024, disk size shown is mixed based 1024/1000 (it is 1024 * 1000 * 1000). The usable size of 1TB disk is 807GiB. b) 10% of aggregate space is reserved for internal use c) By default 5% of remaining space is reserved by snapshots (can be set to 0%) If you compute the result, you get exactly 13109GiB.

aborzenkov · ‎2010-12-11

I just found that neither of storage show shelf or storage show hub would display SATA shelves. What is the right way to see them? On related note - is it possible to deduce from command output whether specific initiator port is connected to IN or OUT and to which shelf? Thank you!

aborzenkov · ‎2010-12-10

I think you will have to recreate all volumes manually. AFAIK volume definition is not kept somewhere on vol0, but in respective aggregate. Keep in mind that LUN serial numbers (if you are using LUN) will most likely change so you may need to reset them as well.

aborzenkov · ‎2010-12-03

If existing aggregate is of type RAID_DP you could convert it it RAID4, that will free extra parity disk in every raid group. Otherwise it is impossible to remove disk from aggregate.

aborzenkov · ‎2010-12-01

No, battery protects NVRAM (or rather it is NVMEM today) to keep its content after power loss; if system is left completely without power for too long, it may happen that write cache is lost.

aborzenkov · ‎2010-12-01

In fabric MC switches used for shelf connection are dedicated to it; it is not supported to connect anything else. At least that is what I have learned so far ☺

aborzenkov · ‎2010-12-01

No, it is not (and is not technically possible).

aborzenkov · ‎2010-11-30

You have to quote quotes for local shell. E.g. using backslash. It is nothing NetApp specific, NetApp gets empty argument which was eaten by your local Unix shell. rsh 2050b options wafl.default_unix_user \"\"

aborzenkov · ‎2010-11-30

You should distinguish between library connection and tape drive connection. Library connection is for control only and usually is done to single system (except in some high available configuration). It does not really matter which system will physically control library. You can connect tape drive to one head and use it to backup via NDMP from second head, but in this case traffic will flow via LAN. Only the first head will be capable to backup directly to tape. But usually libraries have more than one tape drive, so you can allocate tape to each head.

aborzenkov · ‎2010-11-29

Thank you, but it is still unclear (both in fc_iscsi_config_guide_73.pdf and in presentaion) whether I can mix onboard and expansion card of different speed. Both documents state: All expansion adapters should be the same speed So once more direct question - can I mix onboard 4Gb/s target port with expansion 8Gb/s target HBA? Thank you!

aborzenkov · ‎2010-11-29

Were these disks used in NetApp with higher DataONTAP version before?

aborzenkov · ‎2010-11-29

No, it just adds (or removes) additional parity disk.

aborzenkov · ‎2010-11-26

If you extend the aggregate formed by internal disks with disks of the external shelf, then: yes, it will stay raid 4 because this is the aggr option. Or you can change RAID type before adding more disks.

aborzenkov · ‎2010-11-25

Yes to both. You can freely mix sync/async for different volumes (but not for data inside a volume).

aborzenkov · ‎2010-11-25

MetroCluster requires direct SAN connectivity between sites so it is probably out of question here. SyncMirror by itself is not disaster recovery solution (only if used as part of MetroCluster) and have the same limitation – direct SAN connectivity. You could probably look at SnapMirror; depending on your LAN bandwidth and latency you could try using sync or semi-sync mode which will give you close to zero RPO. As you are using VMware, you could consider integrating it with SRM for failover automation.

Re: Verify on SnapVault secondary gets slower every month

Re: Creating a fault tolerant VIF across two switches

Re: SDW 6.x and Space reservation settings on FlexClones just like SDU 4.X ?

Re: Active Directory users integration

Re: Active Directory users integration

Re: File deleted between snapshots - is data still transfered?

Re: How does Unix Mode permissions map to Windows Users in CIFS

Re: Adding new ds14mk2 shelf to FAS2050

Re: Where did the space go?

Re: Where did the space go?

Re: Where did the space go?

How to display AT-FCX shelves?

Re: Restore Volume Structure without Snapmirror

Re: Reassign data disks as spares

Re: Cache data protection

Re: Mixing tape and diskshelf access on the same FC initiator - supported/recommended?

Re: Mixing tape and diskshelf access on the same FC initiator - supported/recommended?

Re: writing cifslog to adtlog.evt manually

Re: FAS3160 active/active cluster and backup with NDMP via single fibre channel

Re: Mixing FC Target Ports Onboard and Expansion

Re: Bad Disk Label

Re: adding shelf / disks to an FAS2020

Re: adding shelf / disks to an FAS2020

Re: NetApp with Disaster Recovery

Re: NetApp with Disaster Recovery