Snap 4100 Slow after Rebuild

[shawnl]

We have a Snap Server 4100 system that had failing HDs and wasn't really being used anymore. (SnapOS 3.4.803) We decided to try and get some more 'life' out of the system, and to try and replaced the HDs in it.

We put in 4 identical brand-new Seagate 120GB 7200rpm drives; and formatted them into a RAID 5.

After it finished building, (I left it overnight, but the log shows it only took a couple of hours) we started backing up some machine images to it; but the transfer speeds were PAINFULLY slow! In windows going to the share, with no other users connected to the snap server, it was only copying the single large image file at about 20% of the 100Mbps link as shown in the windows task manager--about 2.5MBps! If 2 people used it at once the rate dropped down to a mere 6% average (750KBps!)

That seems pretty low for the 'max speed' of this sort of device?

I tried an FTP test, and it is MUCH WORSE! FTP to the 4100 over 100Mbps connection only transfers at 400KBps-500KBps! I can't get an FTP transfer faster than 500KBps! That's only 4Mbps...my home Internet connection is faster than that!

Also, watching the transfers in Task Manager, the connection level 'drops' every 5 seconds or so to almost nothing, and then starts again a second later.

Thinking it might be a network issue, I checked the switch, but there are no errors shown on it. I also swapped the network cables. The same performance was seen from multiple Windows machines to this snap Device. The other windows servers on this switch test at normal expected speeds (they're Gigabit NICs and transfer at 90% of the 100Mbps connection on my computer); and they have similar spec'd drives (3 80GB ATA drives in a RAID 5 for example).

I remember that this snap server used to be 'slow' before but not THIS slow!


Is there any Network settings or things to check on the Snap itself? After replacing the drives, the system retained all it's previous 'settings' (IP address, shares, name, etc), although obviously all data was gone. Should I have reset it to Factory Defaults? What else should I look at? Anyone else have a working 4100? What kind of speeds do you get?

[Phoenix32]

http://forums.procooling.com/vbb/showthread.php?t=13731

Don't anyone read anymore?

[blue68f100]

It takes longer than 2 hr to build the raid5. Takes 20 min to format, 5hr to build raid5.

You may be having a PS problem. Power down and restart and see if it goes into resync.

[Phoenix32]

Originally Posted by blue68f100

It takes longer than 2 hr to build the raid5. Takes 20 min to format, 5hr to build raid5.

resync.

AT THE LEAST...


Also, it will let you access the SHARE after the format is complete, while it is building the "backup drive" parity stripe. While it will let you access it, it will be very slow. This could also be your problem, you might have been accessing it while it was building the parity stripe...

[shawnl]

Originally Posted by Phoenix32

http://forums.procooling.com/vbb/showthread.php?t=13731

Don't anyone read anymore?

I did read that, however the two people with speeds posted there were using a 2200 and a 4000. I was looking for more information about the 4100's abilities.

Originally Posted by blue68f100

It takes longer than 2 hr to build the raid5. Takes 20 min to format, 5hr to build raid5.

You may be having a PS problem. Power down and restart and see if it goes into resync.

I was looking at the logs, and it said the rebuild had completed. *I was a little off on the times however. Looking again at the log, i do see that the format completed in less than 30 minutes at 2:58pm. The RAID "Logical set synchronization done on device 60000" at 8:30pm. It also had time just 'sitting' for 2 days over the weekend after that, before any of the file transfers were attempted.
I was testing it more yesterday, and it had been up and running for 8 days straight at that point. There's no other log messages out of the ordinary.

As for the PS, I suppose it's possible something might be wrong with it, but there's been no drive errors or other log messages... It was 'rebooted' a few times last monday (2 days after being built), and didn't have any issues, but I didn't power it off completely. I'll try that today.



I also noticed that with the debug command "ethernet mtu" the MTU was set to 0. I tried changing it to 1500, but it had no effect on throughput. Also it seems to now be missing the Help files and the JVM. I know it was loaded before, but I guess that was stored on the old HDDs. Could the fact that it still thinks those might be loaded, but really aren't present, be causing any errors? Esp. the JVM? I don't see a way to disable (or any mention at all) now in the interface for the JVM...
What about the FTP speeds? Why would those be so low too?

[blue68f100]

Help files and JVM can be loaded like the .sup files. Are not saved to flashram due to space required.

Writes will be slow do to all of the parity calculations that need to be done. Don't have any bench marks, others may. Reads should be 100baseT limited. Writes may be in the 2 Meg/sec range.

If you shut down normal and start seeing resysn, thats a indications of a HW problem, PS.

[shawnl]

Originally Posted by blue68f100

Help files and JVM can be loaded like the .sup files. Are not saved to flashram due to space required.

Writes will be slow do to all of the parity calculations that need to be done. Don't have any bench marks, others may. Reads should be 100baseT limited. Writes may be in the 2 Meg/sec range.

If you shut down normal and start seeing resysn, thats a indications of a HW problem, PS.

Unfortunately I don't have the .sup files anymore. I've looked through all our storage servers and can't find them from when we upgraded to version 3.4.803 a few years ago. Since Adaptec as since removed those files, it looks like we're out of luck. I'm just wondering if the JVM being setup, but not existing is causing any issues?

It would seem I can get READs from the server at about 40Mbps Maximum (5MBps) with SMB. However I still have no explaination why FTP transfers are so slow at 400KBps?

There were no errors or out of ordinary event after powering off for 10 minutes and turning back on.

[blue68f100]

All of the problems indicates that you may have a drive having problem with seeks, and or bad media. I would run GRC SpinRite v6 on all of the drives. I did this when mine were new. I have 1 drive that has used up 1/2 of the allowed spare sectors. All other checked good. I expect the drive to fail in a short period of time. The Smart Tech built in to the HD will do this on the fly. May be the source of all your problem.

Disable JVM if you do not have it installed.

[shawnl]

Originally Posted by blue68f100

Disable JVM if you do not have it installed.

Where/How do I do this? I don't see options for it anywhere anymore!

[blue68f100]

JVM may not show, without it beng installed. Just make sure you are not tring to do a secure login or www function, which equires it.

I would run a check on the drives like I mentioned earlier. Phoenix32 did this on his 4000 and found a bad drive, his drives were new (OEM). Since Spinrite tried to recover the bad area, it used all of the spares sectors up, then the mfg utility reported an error . I had 1 out of 4 drives use 1/2 of the reserve sector. And these were brand new Seagate drives. So just because there new, does not necessary mean they are good.

Since it is having problems break your array and create a new one. It should take 5+ hr to format the array.

Snap-tech pointed out that drives are only design to last 1 yr. So a normal pc with it up and down time may relate to 3+ yrs use if lucky. Their playing the odds.