I am an IT Professional who lives and works in Melbourne Australia.
This blog is totally my own work.
It does not represent the views of any corporation.
Constructive and useful comments are very very welcome.
RSS Feed – paste this link into your RSS reader software
Hi,
I just started looking into doing PowerHA with XIV Synchronous System Mirror between two sites (less then 100km) and I am looking at any info about it. Do you have any advice in this area you could share with me?
With the best!
MarkD:-)
Hi Mark.
Great blog BTW, I have added it to my links section.
As for the PowerHA, I don’t have anything to hand but I will keep my eyes open.
Hi Anthony,
As the storage expert, I like to inquire about a relatively simple question. Can you expand on how that is accomplished. Does it use the storage manager software to apply the binary to each controller or running an executable? Also, during the upgrade, does the unit have to be downed or if running dual controllers, able upgrade one while running on the 2nd?
Thanks!
Hi Wilson.
Your commenting on the ‘About Me’ page.
Did you mean to comment on a specific post that I created?
Hi,
can you please fix up the rss feed?
https://aussiestorageblog.wordpress.com/feed/
TY!
Hi.
I am confused. The link on the front page is correct (it matches the one you give) which I believe is the feed….
If I use that URL in Google Reader it works without issue (as far as I can tell).
What I am doing wrong?
Hi Anthony,
We had an issue with SVC handling EMC disks because of which we had to take out zoning between EMC and SVC (The issue actually took down the SVC cluster). But the SVC still has the MDISKs, storage pools and Volumes definitions from EMC disk which are in offline state.
Now we plan to delete the definitions from SVC. Any idea will there be any issues if we try to delete these offline mdisks/pools/volumes from SVC?
Thanks
Bharath
Hi Bharath.
There should be no issues doing this although I hope you reported this issue to IBM Service so they can follow up on it.
You will probably need to delete the VDisks with ‘force’ as well as the pool. When you delete the pool it should delete the MDisks.
If any MDisks stil hang around, just do a new MDisk Discovery.
Hi Anthony,
Thanks for the response. Yes we had the issue reported to IBM.
Thanks
Bharath
Hi Anthony, Thanks for the great blog. I’ve have this very odd problem with our XIV’s at work. I’d love it if can get some of your expertise on the matter! :) Do you have an email I can email you at some screeshots
Sure. My email address is anthonyv@au.ibm.com
Hi Anthony,
I am looking for the “traceroute” equivalent for SAN. Do you know about anything like that working with AIX5.3 and above?
Thanks,
MarkD:-)
Great question. From a host perspective there are several applications that can help, like QLogic SanSurfer and Emulex LightPath but I don’t know if there are AIX versions (I dont think so). You are better off working with the switch itself to do this. There are utlities like fcping on both major vendors switches that have improved drastically over time.
Hi Anthony,
Great blog, for our sakes please keep up the great work! There are very few bloggers who seem able to keep focused on the technical (rather than the political) so its refreshing to read your posts.
I have a friend who has a DS5300 issue that you may be able to assist with. The Storage system is located in a remote location in Asia and the local channel have not been able to provide clarity. The problem has been logged with support but they are finding it diffucult to resolve the problem without physical access as they suspect the cabling is incorrect or faulty.
There are 4 FC drive trays connected across all the back end loops plus a SATA drive tray daisy chained off one of the FC trays. Both the FC and SATA trays on the same loop are showing degraded status (2 Gbps instead of 4).
We have ensured that the cabling looks like that recommended in the redbook but do you have a suggestion as to how this system should be cabled to best suit the 5 mixed encolures?
Thanks in advance,
Russell
Hi Russell.
THe cabling is indeed the most likely culprit.
Normally what I do is create a visio for the client and map out the cabling.
Because the DS5100 has up to 8 loops, I normally spread out the enclosures as much as possible, so I ponder why they are all cabled together?
Here is a question/potential blog post. What is the preferred method of performing maintenance on the backend storage behind an SVC? We either move all the vdisks to other storage or shutdown the hosts accessing the vdisk and then do firmware updates, but SVC complains even if there are no vdisks on the pool/controller. Seems like there should be a way to put the controller into “maintenance mode”. Thanks!
Great question.
Right now there is no way to signal to the SVC that a controller needs to go offline, so there is no way to stop the complaints when it does.
The only way to stop the errors would literally be to delete the entire pool (after evacuating its contents somewhere else).
Most clients I work with now use Volume copies (also called VDisk mirrors) to protect data on storage subsystems they are doing maintenance on.
It wont stop the complaints, but doing this gives the highest level of availability.
Hi Anthony,
Great blog, please keep it up. I have a question regarding optimisations on our Informix database as we are currently migrating to a Gen 3 XIV. We use VxVM and I am interested in the performance improvements we can make. We are going to test using different HBA queue depths on the host and different LUN sizes, but I’m unsure at the moment if changing the i/o write size in VxVM and/or striping with VxVM will prove useful. Have you any opinions on this?
Thanks
Hi Jarrod,
Less LUNs and more queue depth both normally have a positive effect on performance. Write size is not such a big issues as all writes are to cache and because XIV uses 1 MiB partitions, pretty well any commonly uses write size will fit inside that. Striping with VxVM is not needed as the data on an XIV is already very widely striped (180 MiB of contiguous logical block addresses is striped across 180 disks – thats about as wide as you can get!).