New Raid 5 setup

[Stuart Bowers]

Hi All,

Just installed Alt-f rc3 abd setup raid 5.

using 2 *  WD Green 3 Tb internal and 1 external usb 3tb Seagate.

The raid took about 1 hour to create.. tho once completed the usb drive was degraded.

I have attached a screen shot and system log for people to review.

I will let the recover action complete and perform some test..

Thanks

PS not putting any important on this as yet.. 

[Stuart Bowers]

Quick update...  Seems this is going to take a while...

Updated the stripe_cache_size to 1024 ... tho does have memory impact(just trying to speed up the rebuild)  lost the web gui too..

any higher and i loose my telnet session...  

I did notice that the chunk size is 512k might be to high? as the system is writing to usb..  64k would be better? thoughts?

Info below.

# cat /sys/block/md0/md/stripe_cache_size

1024

# cat /proc/mdstat

Personalities : [linear] [raid1] [raid6] [raid5] [raid4]

md0 : active raid5 sdc2[0] sdb2[3] sda2[1]

      5859479552 blocks super 1.2 level 5, 512k chunk, algorithm 2 [3/2] [UU_]

      [=>...................]  recovery =  5.3% (158026492/2929739776) finish=2582.6min speed=17886K/sec

      bitmap: 22/22 pages [88KB], 65536KB chunk

unused devices: <none>

#

[João Cardoso]

On Saturday, July 13, 2013 8:42:03 AM UTC+1, Stuart Bowers wrote:

Hi All,

Just installed Alt-f rc3 abd setup raid 5.

Using the disk wizard?

 

using 2 *  WD Green 3 Tb internal and 1 external usb 3tb Seagate.

The raid took about 1 hour to create.. tho once completed the usb drive was degraded.

That is expected and is by design. According to the mdadm (RAID manager program) author, It is faster that way.

 

I have attached a screen shot and system log for people to review.

I will let the recover action complete and perform some test..

Thanks

PS not putting any important on this as yet.. 

The tricky part will be the power up sequence. Start powering up the external disk first, immediately followed by the 323, so the external disk will be available when the internal disks are detected; also, this way the external disk will not be powered up from the USB connector, which does not supply enough current for it. (external 2,5" disks should not have this problem) 

For powering down proceed the other way around, i.e., don't let the external disk be powered from the USB connector.

Be sure that the write-intent bitmap is active, or a lengthy resync will happens after every reboot.

Feedback will be welcome, as there is none regarding true RAID5 operation.

[João Cardoso]

On Saturday, July 13, 2013 11:08:09 AM UTC+1, Stuart Bowers wrote:

Quick update...  Seems this is going to take a while...

Yes, that is expected.

If you search the forum you will find another user that reported a full day (24hours) sync. 

Updated the stripe_cache_size to 1024 ... tho does have memory impact(just trying to speed up the rebuild)  lost the web gui too..

any higher and i loose my telnet session...  

I did notice that the chunk size is 512k might be to high?

That's is the default. If you know that your usage pattern will benefice from smaller values, go for it (and reporting back your experience will be welcome; users experience on RAID5 on PCs or servers is abundant, but not on low memory systems)

[Stuart Bowers]

ola João,

Thanks for the info.

Yes I used the disk wizard,  found a little bug in the web gui  see the below Done status on the raid recover.

the real info is.. 

 cat /proc/mdstat

Personalities : [linear] [raid1] [raid6] [raid5] [raid4]

md0 : active raid5 sdc2[3] sdb2[0] sda2[1]

      5859479552 blocks super 1.2 level 5, 512k chunk, algorithm 2 [3/2] [UU_]

      [=======>.............]  recovery = 37.1% (1088758992/2929739776) finish=1920.9min speed=15972K/sec

      bitmap: 15/22 pages [60KB], 65536KB chunk

unused devices: <none>

[João Cardoso]

On Sunday, July 14, 2013 3:53:15 AM UTC+1, Stuart Bowers wrote:

ola João,

Thanks for the info.

Yes I used the disk wizard,  found a little bug in the web gui  see the below Done status on the raid recover.

thanks, seems to be a kernel bug, see issue 140

[Stuart Bowers]

Hi All

Quick update.. The raid recover is now complete. took just over 2 days.(for my setup)

web-gui

Started a SMB file copy..(1.06gb file) to dns-alt-f(rc3)

CPU looks like it peeks out

Using top via telnet

Copied same file to DND-323 stock 1.04(non raid volume)

Preformed a reset(power) via Webgui..  no raid issues..  all mounted fine.. 

Preformed a power off via Webgui..  no raid issues..  all mounted fine.. 

Will continue testing and provide some more updates.  Thanks for this great firmware João

[TJ]

I was wondering if you could satisfy my curiosity, and perform a dns 323 to pc transfer of that same 1GB file. I assume you're using SAMBA?

T.J.

[Stuart Bowers]

Hi TJ,

Wow yes much better copying from the dns-323

Also did try this too

seems a slight improvement on transfers(uploads only) extra 1 mb p/s

socket options = TCP_NODELAY IPTOS_LOWDELAY SO_RCVBUF=65535 SO_SNDBUF=65535

Thanks

Stuart