StarWind iSCSI SAN Version 8.0 Beta-3

[sbozhko]

3. Extend partition after that - all OK

Extend partition after that - all OK

20140131_114157.png (47.35 KiB) Viewed 8936 times

[sbozhko]

I got this error on two different systems:

1. Operating System Microsoft Windows Server 2012 R2 Standard 6.3.9600.16452 (Win8.1 RTM)
CPU Type 2x OctalCore Intel Xeon E5-2680, 1200 MHz (12 x 100)
Motherboard Name Dell PowerEdge R720xd
System Memory 131026 MB (DDR3 SDRAM)
Storage Controller PERC H710P Mini
Disk Drive DELL PERC H710P SCSI Disk Device (185 GB) (RAID 10)
Disk Drive DELL PERC H710P SCSI Disk Device (19268 GB) (RAID 10)

2. Operating System Microsoft Windows Server 2012 R2 Standard 6.3.9600.16452 (Win8.1 RTM)
CPU Type 2x QuadCore Intel Xeon E5335, 2000 MHz (6 x 333)
Motherboard Name Supermicro X7DB8 (3 PCI-E x8, 3 PCI-X, 1 SEPC, 8 FB-DIMM, Video, Dual Gigabit LAN, SCSI)
System Memory 24576 MB (DDR2-667 Fully Buffered ECC DDR2 SDRAM)
Storage Controller Adaptec RAID 6805/6805Q
Disk Drive Adaptec Array SCSI Disk Device (278 GB) (RAID 10)
Disk Drive Adaptec Array SCSI Disk Device (2399 GB) (RAID 10)
Disk Drive Adaptec Array SCSI Disk Device (2796 GB) (RAID 5)

[fbifido]

Hi,

just installed v8.0.6312(beta-3)

Tried to create a 1024GB thin LUN, but failed with error message, see pic below.

http://i44.tinypic.com/t69w93.png (full image)


http://oi44.tinypic.com/t69w93.jpg (RAW image)

I have 12GB RAM, about 2G is been used.
I have 4TB SAS, about 2TB free.


I tried creating a 10GB LUN, I did not get that error.
I tried 100GB LUN, did not get that error.
tried 500GB LUN, no error
tried 900GB LUN, no error

Tried 1024GB LUN, error
Tried 1TB LUN, error.

Installed v8.0,6345, created a test LUN of 500GB , I tried 1024GB but still got that message.
anyway I add the new lun to esxi 5.5, then restart my esxi server as well as my SAN.
now esxi will not auto add my old FLAT lun bact to its storage, I have to manually add it back., but it did auto add the test lun (LSFS).

I then deleted the FLAT LUN, and tried to create a new LSFS, I tried 2048GB then 1024GB but the message is still the same.
with only one 500GB LSFS LUN, StarWind iSCSI SAN service is using 3GB of RAM from a total of 12GB, the total use memory is 4.9GB of RAM.

------- Please note that this is a BUG in the LSFS code ------------------

With only 7GB of RAM remaining I can only create a 627GB LSFS LUN. This is very vexing
and when you accept the LSFS surjection and created the 627GB LUN you will not get any write-back caching, which is the part that gets rid of the blender effect problem.
Anyway it created the LUN, then I try to created another LUN 1024GB LSFS, but this time is says I can only create a 638GB LUN, I created it, then tried another again this time it says 659GB LUN.

So I can create a lot of small size LSFS LUN but not a big size one. WHY IS THAT ?

What have change from v8 beta 1 to v8 beta 3, that remove the ability to create a virtually any-size "THIN" LSFS lun? in the mean time can you get write-back-cache, dedup & auto defrag to work on THICK LUNs.



So I see now two main problem I am having with the iSCSI SAN v8 beta 3
1. I can't create a large size LSFS LUN.
2. The problem of LUN disappearance from esxi 5.5 storage when I update iSCSI san and restart esxi.

[fbifido]

Better use it AS IS (LSFS but no deduplication) and use flash as a flash cache. More IOPS.

just saw this info from a post and tried it.

I tried to create another LSFS LUN, but this time I did not enable dedup, and it says the max I can create is 912GB LUN.

I only want to create 2 LSFS LUN out of my 4TB StorageSpace POOL

1 2TB LSFS LUN for VDI & VM
1 2TB LSFS LUN for Documents & e-Mails
1 NFS for ISO & Templates using the remaining space save from the dedup.

The reason I make the LUN large is because, ones the iSCSI reports to ESXi that the 2TB LUN are full, that does not mean the 4TB Pool itself is full, I can then create another LUN, say 1TB, to move some of the lease use VM to it, so make sure you have LUU to LUN copy with dedup intact.

[Bohdan (staff)]

2. The problem of LUN disappearance from esxi 5.5 storage when I update iSCSI san and restart esxi.

Are you talking about v6 to v8b3 update? v6 to v8b3 must work fine.
v8b1, v8b2 update to v8b3 may cause the LUN disappearance.
Please make sure this is not v8beta1 (or v8beta2) to v8beta3 update.

As for the LSFS sizes, my colleagues are preparing the answer and they'll post it here ASAP.

[Vitalii (staff)]

When the new LSFS device is created it does not use all needed memory. Memory usage grows as device is filled with data.
Maximum size restriction is based on the assumption, that there should be enough memory to support 100% full device. Until that, it does not use additional memory, so it is still available for the system.

That is why, you are able to create multiple devices of smaller size, but are not able to create single large device.

Vitalii

[Bohdan (staff)]

I got this error on two different systems:

1. Operating System Microsoft Windows Server 2012 R2 Standard 6.3.9600.16452 (Win8.1 RTM)
CPU Type 2x OctalCore Intel Xeon E5-2680, 1200 MHz (12 x 100)
Motherboard Name Dell PowerEdge R720xd
System Memory 131026 MB (DDR3 SDRAM)
Storage Controller PERC H710P Mini
Disk Drive DELL PERC H710P SCSI Disk Device (185 GB) (RAID 10)
Disk Drive DELL PERC H710P SCSI Disk Device (19268 GB) (RAID 10)

2. Operating System Microsoft Windows Server 2012 R2 Standard 6.3.9600.16452 (Win8.1 RTM)
CPU Type 2x QuadCore Intel Xeon E5335, 2000 MHz (6 x 333)
Motherboard Name Supermicro X7DB8 (3 PCI-E x8, 3 PCI-X, 1 SEPC, 8 FB-DIMM, Video, Dual Gigabit LAN, SCSI)
System Memory 24576 MB (DDR2-667 Fully Buffered ECC DDR2 SDRAM)
Storage Controller Adaptec RAID 6805/6805Q
Disk Drive Adaptec Array SCSI Disk Device (278 GB) (RAID 10)
Disk Drive Adaptec Array SCSI Disk Device (2399 GB) (RAID 10)
Disk Drive Adaptec Array SCSI Disk Device (2796 GB) (RAID 5)

Thanks! We reproduced the problem. Working on the fix. I believe the new build will be available in 1-2 days.

[fbifido]

2. The problem of LUN disappearance from esxi 5.5 storage when I update iSCSI san and restart esxi.

v8b1, v8b2 update to v8b3 may cause the LUN disappearance.
Please make sure this is not v8beta1 (or v8beta2) to v8beta3 update.

it was v8b3(8.0.6312) to v8b3(8.0.6345)

[fbifido]

When the new LSFS device is created it does not use all needed memory. Memory usage grows as device is filled with data.
Maximum size restriction is based on the assumption, that there should be enough memory to support 100% full device. Until that, it does not use additional memory, so it is still available for the system.

That is why, you are able to create multiple devices of smaller size, but are not able to create single large device.

Vitalii

Sorry but, don't scare people like that.

Please explain this a little bit more please, in cartoon form so I can understand:

I create a LSFS LUN, it checks the free memory I have and then calculate what is the max LSFS I can create and if what I want is greater it lets me know, and make a recommendation, now this suggestion is base that if this LUN became full then all the memory in my server will be fully use (yes/No)?
What if I had created 2 LUN, does that means the if one of them is at 100% the system will run out of memeory? (Please don't tell me this is like that FreeBSD dedup problem)

Can I use windows server 2012 R2 dedup & disable iSCSI dedup on the LUN in starwind software.

What I understand with default settings:
iSCSI SAN Services use 3GM of RAM
each LUN need 128MB RAM for Blender fixing (L1 cache) and adds 20MB RAM to iSCSI SAN Services if dedup is disable.
if dedup is enable I don't know how mush RAM it will add to the iSCSI SAN Services.

So when data is received for a LUN it is cache to L1 of that LUN, where it is defragmented & dedup(if enabled), a responds is send to the data sender that it was written, then if L1 is not full it waits for more data or write out to disk.:
1) after getting a few more data, do defrag & dedup, then write it to permanent disk.
1) after getting a few more data, do defrag & dedup, then wait for more data, until cache is full, then write out to permanent disk.

[Bohdan (staff)]

I got this error on two different systems:

Please re-download the beta build. It should solve the problem.
http://starwindsoftware.com/tmplink/StarWindBeta.exe

Note: This build also fixes LSFS update procedure (datastore disappearing), so updating THIS (6383) build to itself or to the next (future) builds will not cause datastore disappearing.

[turkina]

Is it still a requirement to only update to the latest beta build from v6 and that updating to the latest beta build from a previous beta build could fail?

[Vitalii (staff)]

I create a LSFS LUN, it checks the free memory I have and then calculate what is the max LSFS I can create and if what I want is greater it lets me know, and make a recommendation, now this suggestion is base that if this LUN became full then all the memory in my server will be fully use (yes/No)?

Yes

What if I had created 2 LUN, does that means the if one of them is at 100% the system will run out of memeory? (Please don't tell me this is like that FreeBSD dedup problem)

Yes

About the cache. Write-back cache is flushed to disk in two cases:
1) when dirty cache block has not been read or written during five seconds and there is no disk load at the current moment.
2) when new write operation arrives, target address is not in cache and all cache blocks are dirty. Oldest block is flushed to disk to make space for the fresh data.

After that, blocks are processed by deduplication engine and is written to disk. The deduplication engine has its own cache (usually 4 or 8 MB) and it is written when it is full. It is close to the first option you mentioned: data is written to permanent disk when there is 4 MB of cached data (not the value you chose for L1 cache).

[Bohdan (staff)]

Is it still a requirement to only update to the latest beta build from v6 and that updating to the latest beta build from a previous beta build could fail?

It is not a requirement. It is warning. Due to changes/fixes related to vendor/product, serialId and eui-64 identifier there is possibility of datastore disappearance and and necessity to manually re-add them.
v6 to v8b3 (6345 or later) should work fine for all device types.
LSFS device update should work fine in v8b3 (6383 or later).

[turkina]

Is it still a requirement to only update to the latest beta build from v6 and that updating to the latest beta build from a previous beta build could fail?

It is not a requirement. It is warning. Due to changes/fixes related to vendor/product, serialId and eui-64 identifier there is possibility of datastore disappearance and and necessity to manually re-add them.
v6 to v8b3 (6345 or later) should work fine for all device types.
LSFS device update should work fine in v8b3 (6383 or later).

What if we're not using LSFS and only image files? Then is an upgrade from beta2 to beta3 safer?

[sbozhko]

I got this error on two different systems:

Please re-download the beta build. It should solve the problem.
http://starwindsoftware.com/tmplink/StarWindBeta.exe

Thanks! With the new build I have successfully created LSFS volume on a large partition!
But I noticed a small problem. When I create a L2 cached volume, Add Device Wisard is always named cache img-file "L2Storage1.img", regardless of what I have in the Wisard (swdsk-file named correctly). Respectively, it is impossible to create the next volume with L2 cache on one partition (without manual .swdsk corrections).