The IBM Storwize V7000 has a new stable mate: IBM’s hot new seller, the Storwize V3700. I recently got a chance to try one out and I liked what I saw. I have always tried to share useful information on this blog, so here are four things you may find useful about IBM’s new little midrange storage offering:
Node Canisters
The Node Canisters (Controllers) are side by side and both right way up. I really like this change. Hopefully all future models will follow this pattern and avoid upside down components. One thing you will spot from the picture is that the Fibre Cards are optional. What you might think are Fibre Ports in this picture are actually SAS ports. The fibre card goes where that large black square is on the right hand side of each canister.
Environmentals
The Storwize V3700 can report power consumption and operating temperature via both the GUI and CLI. This is a great extra piece of information.
Being able to get this information via CLI is also critical as it allows you to script it for those shops where rack power consumption is constrained so check out the lsenclosurestats command.
IBM_2072:Cluster:anthonyv>lsenclosurestats enclosure_id stat_name stat_current stat_peak stat_peak_time 1 power_w 124 125 130128230402 1 temp_c 19 19 130128230707 1 temp_f 66 66 130128230707
License Tab
I looked for the license tab…. but there isn’t one! This is because Flashcopy is included, external virtualization ( as a migration tool) is included and remote copy is not possible. This makes for very simple purchasing; all you need to do is decide what disks, RAM and adapters you want. Nice!
I did find one (tiny) bug that is easily corrected, but is stealing 40 MB of your cache! If you display the bitmap memory, you may find 20MB dedicated to remote copy, despite the fact that you cannot create remote copies.
IBM_2072:Cluster:anthonyv>lsiogrp 0 id 0 name io_grp0 node_count 2 vdisk_count 3 host_count 1 flash_copy_total_memory 20.0MB flash_copy_free_memory 20.0MB remote_copy_total_memory 20.0MB remote_copy_free_memory 20.0MB mirroring_total_memory 20.0MB mirroring_free_memory 20.0MB raid_total_memory 40.0MB raid_free_memory 39.3MB maintenance no compression_active no accessible_vdisk_count 3 compression_supported no
You can easily correct this by running the following command that drops that bitmap to zero. You can run this command at any time, there is no risk in doing so. You will get 40MB of cache back (20MB per node canister).
chiogrp -feature remote -size 0 io_grp0
WWPN Determination
I spotted two interesting things about the WWPNs for the Storwize V3700 ports. Firstly IBM has broken with the 1,2,3,4 pattern we found with Storwize V7000 and gone to 04,08,0C,10. Frankly this is not a big deal and given the Node Canisters are side by side, it is just a case of knowing the pattern. The WWPN is based on: 50:05:07:68:03:YY:xx:xx where xx:xx is unique for each node canister and the YY value is taken from the port position as per the image below. I suspect these values may go up to 05, 09,0D,11 over time as they exhaust the serial number range possibilities of 00:00 to FF:FF
I did spot what I think is a great new command in V6.4.1 that also lets you display the WWPNs. It is lsportfcid. Try it out on your machine.
IBM_2072:Cluster_1:anthonyv>lsportfcid fc_io_port_id port_id type port_speed node_id node_name WWPN nportid status attachment 0 1 1 fc 8Gb 1 node1 5005076803040046 010500 active switch 1 2 2 fc 8Gb 1 node1 5005076803080046 010000 active switch 2 3 3 fc N/A 1 node1 50050768030C0046 000000 inactive_unconfigured none 3 4 4 fc N/A 1 node1 5005076803100046 000000 inactive_unconfigured none 6 1 1 fc 8Gb 2 node2 5005076803040047 010400 active switch 7 2 2 fc 8Gb 2 node2 5005076803080047 010100 active switch 8 3 3 fc N/A 2 node2 50050768030C0047 000000 inactive_unconfigured none 9 4 4 fc N/A 2 node2 5005076803100047 000000 inactive_unconfigured none
I did spot one thing when displaying the FC ports in the GUI. They are currently listed back to front, just something to be aware of:
So are you running a V3700? How is it working out for you?
Hey Anthony,
Another great post and very timely !
Our stack of v3700 turned up on site this week and we are currently setting it up to work behind the SVC that we have.
Really looking forward to putting into production use.
Thanks again,
Christian.
The changes in the FC WWPNs are actually to accommodate FCoE ports, which show up in the lsportfc command alongside FC ports if you have the 10Gb/s Ethernet option. The lsportip command shows IP addresses used for iSCSI.
Nice…. can you tell me we would decode the FCoE ports WWPNs?
Pingback: How to compress 1 Gb data to 10 mb - Post2sale
Pingback: Storwize V3700 – First impressions « Storage CH Blog
any issue with turning off flash copy as well to recover another 40MB? if sure that you won’t be taking any flash copies?
Go for it. It’s a good idea
You can turn it back on without an issue if you need to later
v3700 drive fw pack was out, updated using the command line (from one of your v7000 posts). is there another way / GUI option to do this? thanks.
Not as yet AFAIK
Does it support Remote mirroring also?
No it doesn’t
Then the cover page of sg248107.pdf looks like lies…
I just spotted there is a statement of direction to support it. Maybe I have old info
so does “mirroring_total” refers only to metro/global mirror? then “remote_copy_total”? what;s the difference?
Remote copy is for metro and global mirror which v3700 cannot do
Mirror total is for vdisk mirroring which is local copy for having two copies for a vdisk
thanks, seems in general use you would only need to leave raid.
v3700 iscsi with windows 2008 native mpio, detects opt and non opt paths (non opt being paths to non preferred controller for writes?) so RR (subst) over RR is preferred?
Correct you would detect preferred and non preferred paths. On a per volume basis it send IO to only preferred paths
is flow control supported on v3700 iscsi models? other than connecting to a fc enabled switch port, any configuration required on v3700 ports? thanks.
Hi, Anthony,
I encountered a problem with v3700: when I try enable write-caching-policy in OS (Win2012, Win2008R2) – I get an error “windows could not change the write-caching setting for the device”. Is it really bug or its feature? For DS35xx i haven’t this problem – i can to enable write-caching-policy in OS. LUNs from v3700 are mapped to servers via FC switches (IBM 2498)
I don’t believe that Windows can set write-caching on or off for SAN Storage volumes. I think that setting only applies to internal SAS/SATA drives rather than externally presented LUNs.
Hi Anthony.
We has IBM Storwize v3700 at configuration: 1 Controlled enclosure + 2 Expansions. All enclosures consist of 36 disk, SAS, 3 Tb, 7200 RPM. v3700 connect to SAN via iSCSI. v3700 works very slow: CPU Usage >50%, Disk Latency > 20 ms. It`s normal behavior for this system ?
That sounds very curious, especially the CPU usage.
I would place a service call with IBM.
I have v3700 and ds4700 as SVC backend storage. Tried to use add mirror copy to one of the vdisk (primary-ds4700) and the mirror copy (v3700) with size of 200gb.
It took almost 17 hours for the sync to complete. Is there any parameter need to be adjusted or it is behaving as expected?
The default sync rate is 2Mbps
You can increase it all the way to 64Mbps
This is done by CLI with chvdisk command
Cheers for the info. Will try it out.
Hi, I saw that this Storage support now 6GB Host Attach.. but on SSIC we did not find this compatibility to HBA 6GB SAS.Did you note that?
Hi Anthony,
Currently we are using
DS4700 Express 4GB FC
4GB cache
16* 146GB 4GB FC Disk
SAN Switch 4GB.
I m planning to migrate to
V3700 8Gb FC
8GB Cache
24*300GB 6GB SAS Disk.
SAN Switch 8Gb.
Will i get significant oracle database performance improvement?
Hi Anthony,
sorry .. following is the current config,
Currently we are using
DS4700 Express 4Gbs FC
1GB cache percontroller
16* 146GB 4Gbs FC Disk
SAN Switch 4Gbs
Hi Anthony,
Just configured a new V3700 with direct attach to the 8 Gbit FC adapters on a new p720. No FC switch involved. The ports from the p720 are not discovered using the rescan button in the GUI. Added the ports’ WWPN but still offline. Latest firmware on the V3700 and the HBAs. The 720 is running Virtual I/O Server 2.2.2.2, which owns the HBAs.
Should I install the Host Attachment Kit / SDDPCM for the VIOS? There’s none available for the VIOS / AIX combination, but there is for a directly-connected AIX host.
http://pic.dhe.ibm.com/infocenter/storwize/v3700_ic/topic/com.ibm.storwize.v3700.710.doc/svc_fibrechannel_cover.html
SDDPCM is available here:
http://www-01.ibm.com/support/docview.wss?uid=ssg1S4000265
Even direct connected you still need MPIO set up, so I would definitely install a PCM.
I don’t see IBM Host Attachment for SDDPCM on AIX on page:
http://www-01.ibm.com/support/docview.wss?uid=ssg1S4000203
as prerequisite for SDDPCM.
Does it mean it is needed for V7000 but for V3700 not?
Its needed for both
Non-sequitir does anyone know how to change the FC port_speed on a v7000 mine shows 2 GB despite 8 Gb SFPS attached to 8 GB SFP switch (brocade 8 Gb 1u chassis) is at auto-negotiate. I read a blurb on change in SVC at front panel but not for v7000 or v3700. Thanks
Is it really on 2 GBps?
I think that is a legacy status message which is not actually the truth.
Back in the very early days of SVC we would set the speed of an entire node with this setting. I think it still displays that value, but does nothing with it.
That’s likely correct Anthony, the 1Gb/2Gb setting still exists for the 2145-8F2 node that is still supported. From back when FC hardware wasn’t able to auto negotiate very well.
To check the real port speeds you can use “lsportfc” on the CLI.
Spot on Alan.. Jerry where are you seeing ‘2Gbps’ ? Run lsportfc as per Alans advice.
Starting with FOS levels v6.2.0, v6.2.0a & v6.2.0b, Brocade introduced arbff-arbff as the new default fillword setting. This caused problems with any connected 8Gb SVC ports and these levels are unsupported for use with SVC, Storwize V7000 or Storwize V7000 Unified.
http://www-01.ibm.com/support/docview.wss?uid=ssg1S1003699
Anthony, have you had chance to test the Storwize v3700 alongside VMware & Veeam? We have three units (all 1Gbps iSCSI) acting as datastores for ESXi 5U2. They appear to work OK with active full backups (using DirectSAN attach), but come to an absolute crawl during incrementals. Veeam 6.5 P3 shows 547KB/s for the data rate during this. Storwize firmware is 7.1. Thanks.
hi Anthony,
Just installed two Storwize v3700 connected to Pureflex chassis via IBM SAN24B-4 switches. However we are having some problems during the installation. I mapped one volume from each V3700 to a x440 compute node having Fc 3171 2 port 8Gbps adapter. Multipathing is configured on the x440 compute node running Win Svr 2012 Std Edition. The initialization of the disks goes fine but when we try to convert the disks to Dynamic, the Disk Management Console hangs and the below error is reported in Event viewer:
The IO operation at logical block address # for Disk # was retried.
Any idea for above issue?
Please advise
If you leave the disks in basic mode can you read and write files without issues?
Have you installed SDDDSM
http://www-01.ibm.com/support/docview.wss?uid=ssg1S4000350
DIdnt try with basic disks.
But when i try with small volume like 2gb it works. but when i try with large volume like above 100Gb, the problem occurs.
Yes i installed SDDDSM 2.4.3.4-4. I also updated the firmware version of the HBA on the compute node to 9.11. Before upgrading the version of 7.0 and was gettting “Reset to device, \Device\RaidPort0, was issued” in event viewer.
Do you think its an issue with the mulitpathing?
Disk size at those values should make no difference. Some strange driver issue going on.
Would love to see if basic disks don’t have this problem.
If you have SDDDSM installed then open up that app (it will be a command prompt) and run:
datapath query device
qatapath query adapter
Are you getting errors?
Would love to see the output.
Hi Anthony,
Here you are:
C:\Program Files\IBM\SDDDSM>datapath query device
Total Devices : 2
DEV#: 0 DEVICE NAME: Disk0 Part0 TYPE: 2145 POLICY: OPTIMIZED
SERIAL: 600507630080824AB80000000000000F LUN SIZE: 400.0GB
============================================================================
Path# Adapter/Hard Disk State Mode Select Errors
0 Scsi Port1 Bus0/Disk0 Part0 OPEN NORMAL 0 0
1 Scsi Port1 Bus0/Disk0 Part0 OPEN NORMAL 246 2
DEV#: 1 DEVICE NAME: Disk1 Part0 TYPE: 2145 POLICY: OPTIMIZED
SERIAL: 60050763008082523800000000000010 LUN SIZE: 400.0GB
============================================================================
Path# Adapter/Hard Disk State Mode Select Errors
0 Scsi Port2 Bus0/Disk1 Part0 OPEN NORMAL 178 0
1 Scsi Port2 Bus0/Disk1 Part0 OPEN NORMAL 0 0
C:\Program Files\IBM\SDDDSM>datapath query adapter
Active Adapters :2
Adpt# Name State Mode Select Errors Paths Active
0 Scsi Port1 Bus0 NORMAL ACTIVE 246 2 2 2
1 Scsi Port2 Bus0 NORMAL ACTIVE 178 0 2 2
C:\Program Files\IBM\SDDDSM>
i cant connect to Storwize 3700, no ip address configured. Any way to configure another ip address
Have you tried seeting the IP addres with the USB stick?
Anyone tried these on sas switches?
I got 2x20MB memory from remote_copy_total_memory. But where does v3700 use it? Is it good and possible to increase raid_total_memory or flash_copy_total_memory?
Hi.
The only reason to increase memory is if the machine asks you to do it.
Adding more RAID memory wont make your machine perform better.
Its only used for tracking bitmaps, so it uses as much as it needs.
You can reduce memory to get rid of preallocated memory for functions you dont use.
SO it you are not doing remote copy, get rid of that memory (drop it as low as it can go, which I think is 1 MB).
But having said that… you are getting back a trivial amount of cache memory.
Would anyone know the CPU currently being used on V3700?
My 1 year old machine has 2 core Intel(R) Celeron(R) CPU G530T @ 2.00GHz in both canister.
hello
i have 3700 with only 5 ssd 400gb 12gbs configur in raid 0
connectect with qlogic fiber optic 8Gbs
i have slow prefrences
when i test with parkdale
i have write speed 144 mb/sec random read 16MB/s and only 4106 iops
and read at 282MB/s random 47MB/s 12030 iops
and when i connect it direct attach my ssd to an array to my server
i get
write speed 480MB/sec random read 85MB/s and 22000 iops
and read at 450MB/s random 75MB/s 18500 iops
can some one pleas help me i have no idea why
How did you tested it?
parkdale batch mark
and
in sql query that take like 15 sec to run on the direct attach
and when the ssd is on the storage it took like 50 second
just a test
same disks
Dear Sir,
I have found your blog on searching on google for problem of sync very slow between our two same storage (Production) and (DR Site) configuration IBM Storwize V3700 with (Latest firmware version 7.5) . Grater than 1 or 2 GB sync is very slow. I red all your questions and reply above blog . So what could you suggest for it ? one person said in query it’s sync default rate is 2 MBps that’s why problem create ? I have step as per documentation as per your view any suggestions please give reply. I am awaiting your reply.
Great Article
can any expert help me in my problem .
i have Head office and DR Site storwize v3700 i can ping both Ethernet ports .but when i am creating partnership it gives error that unreachable ip address .
your replies will be highly appreciated !
will answer my own Question LOLx !
i changed both layers to Replication and it start working .
Excellent article! I’m Planning to deploy Easy Tier. Do you know if any SSDs not on the list of supported drives (http://www-01.ibm.com/support/docview.wss?uid=ssg1S1004229) can be used with Easy Tier on the V3700, eg Intel SSDs, Would the V3700 system still be supported? Thanks.
If you mean drives not sourced from IBM then they would not be supported….