DSM-G600, DNS-3xx and NSA-220 Hack Forum

How about a search?

In the early firmware versions (I believe prior to 1.02b) the ext3 file system was offered as a choice in the format menu, and seems to have been linked to reports of massive data corruption - I do know that after the option was removed the reports seem to have stopped, I don not know if any work was done on the ext3 implementation in the ensuing firmware revisions - 1.03/1.04/1.05 - but if you're offering to "guinea pig" it, then go brave.

I am now "guinea pigging" EXT3 on 1.05 in a raid 1 config.  Other than the reformat to EXT2 message that pops up when I log into the Web config, it appears to be functioning without a hitch.  I have copied roughly 9000 files totaling 38GB of various types of media content.  Directory structure and playback appears to be working correctly.

Anyone have first hand experience on what problems to expect and how easily to reproduce?  I plan on running a mass file compare later this week after loading up a bunch more data, break/rebuild raid, and doing a handfull of standard and nonstandard shutdown, powerloss scenarios.

Cheers!
BSP

Have you tried a restore or a compare of the stored data?

One of the things I do for peace of mind is run a verify as a part of the backup.

bspvette86 wrote:

I've been a hack since home computers came with LED displays, were backed up on cassette recorders, upgrading from 1k to 4k of memory was a big deal, and modems had acoustic couplers...  Boy was it hard to clean up after the new years party when we used the paper tape punch holes for confetti!
BSP

Gentlemen - I think I've been officially replaced as the oldest forum member - champagne, cigars and a ticker tape parade are in order here.

BSP - so what was your first system?

Dlink wrote:

So does this mean BSP will start making posts about Raid and BACKUP to fill in your Role fordem? hah.

Don't go sassing your elders now

bspvette86 wrote:

PS.  Unlike Fordem, I use Raid-0.  But similar to Fordem, I use protection!

condoms?

Off topic and not something that you shout about really but . . .

My first computer was a mainframe with card input at the tech college.  Seem to remember it was an IBM 4130.  Ran my programmes twice a week and got kicked off if it failed to finish within two minutes - FORTRAN and Algol.

When I went to university we had a state of the art Mini.  A PDP 11/8 with one k of core store.  It was a multi-user jobbie with five terminals, and paper tape output.  One of the guys used to programme it in assembler - we were not worthy. 

And that nice concurrent CPM.  What ever happened to that?  Used it on my Apricot for a while.  And the ZX81 . . . with the new fangled cassettes . . .

Before you ask, yes I still use WordStar version 7.0 - probably the best of at least 14 word processors I have had to learn (and love).  Keep your word for windows (doesn't deserve capitals).  F3, F3, F3, yes ring any bells for any of you?  I remember WPS plus was one of the very worst. 

Dbase 2 and Dbase 3 and Condor and FoxPro and  . . . I still think SuperCalc is the best spreadsheet.  What is all that having to put a plus sign in front of a formula thing all about?  Pixx poor programming if you ask me - oh yeah, it's a miscosoft product - a rarity in that you can kill most of its programmes from the keyboard (!) without really trying. 

Never really got into that Unix stuff.  The boys were very - only idiots don’t understand UNIX - which rather put me off. 

Now I'll just sit here and wait for the - that's nowt lad, when I was a boy  . . . here it comes !

Biscotte

Quick Update:
Been running 48 hours now with no problems.  Have written over 500 GB of data with no data corruption or directory problems.  (the types of errors I found googling)

The one problem I have found so far is that when FTPing an 89GB file, the DNS-323 closes the connection at the two hour mark.  Have tried the same send twice now and it bombed at two hours both times.  The same file transferred and verified via SMB fine...

Testing continues...

Cheers!
BSP

I have been thrashing this thing to death the last coupled days and have yet to find any data issues.  I have written and read close to 1 TB of data, filled the disks to 100% and got the email alert, done large local copies and tars within the unit from telnet,  let the power management sleep the drives, created large numbers of small files, mixed sizes of files, a few really large files, created large directory trees,  used the various services (SMB, UPnP, iTunes, FTP) from various devices and OSes without any reboots. 

At this point, I am going to do some bad things like pulling the power, popping a drive out to break the RAID-1, rebooting in the middle of file copies, etc... to see if that has any impact on anything other than the file that is being updated in the middle of the failure.

Does anyone have any other specific tests they would like to see run?

Cheers!
BSPvette

Last edited by bspvette86 (2008-05-22 04:32:39)

jdoering wrote:

Personally the risk I see here is that unless there's some science to why EXT3 was a problem on the DNS-323 in the first place; that a random low-volume test won't tell us much about whether or not that problem went away.

EXT3 has been (/is) a very reliable filesystem in many production *nix servers long before the DNS-323 existed... so intrinsically I don't see why it should have been a problem for the 323 in the first place. (Note that in addition to being much newer to computers than some of you in general; I'm also newer to the DNS-323 than anything pre-1.04.) But reports are that it was a problem... so unless there's been active work to UNDERSTAND and fix it; how much confidence do you have that low-volume single-user testing validates an improvement? (Any single user testing is relatively low-volume compared to massive parallel efforts that tend to expose "random" bugs).

That said; I typically like to live on the edge and with technology ask why not rather than why... so in spirit I agree with the experiment; but in this case I do wonder about bang-for-the-buck.

-Jeff

Jeff,
I understand your concerns and my next line of testing includes going back to version 1.02 of the firmware to confirm that I can duplicate the data loss issues.  Please note that between 1.02 and 1.03 the core release of the linux kernel used in the DNS-323 made a version change.  It would be interesting if the problem happens in 1.02 but not in 1.03.  That would have made the removal of EXT3 unnecissary.

My bang-for-the-buck is general improvement of the product.  I really really really hate to see products loose functionality rather than being fixed correctly. 

Cheers!
bspvette

Don't treat the symptoms, fix the root cause!!!

Well, that didn't take long to break with firmware 1.00.  Started with a complete reset to defaults after backdating the firmware, two "new" (subtitute: non-partitioned) disks,  Formatted to EXT3 and Raid-1.  Kicked off a copy of my pictures folders and when I came back, massive corruption,  directory entries duplicated, etc...     Raid 1 kept loosing sync and rebuilding.

Moving on to Firmware 1.01 to do the same test....
  Duplicate directory entries,  corruption.  Pretty quick!

Moving on to Firmware 1.02 to do the same test.....
  Duplicate directory entries, Corruption. 

Moving on to Firmware 1.03 to do the same test.....
  So far so good....


Cheers!
bspvette

Last edited by bspvette86 (2008-05-24 15:05:18)

I have wrapped up my 1.03 testing of EXT3 after two days without finding any corruption.  I am skipping tests on 1.04 and going back to 1.05 for all remaining testing.

Cheers!
bspvette86