Can't create/modify users, can't create/modify shares, can't copy big files, ...

[hawihoney]

Seems that I don't have any luck with RC11. After moving my two original LimeTech machines from 4.7 to 5.0RC11 I do experience a lot of problems. I painfully followed the migration steps.

 

unMenu is installed with the packages openssl and openssh (SSH doesn't work either, so I'm working with Telnet when connecting to the machines). SimpleFeatures was installed but has been removed the other day.

 

 

* Tried to modify the include disks field on the share page. Whenever I hit "Apply" I'm back to that same page with the old value filled. No hint in the syslog. Every change I do is simply ignored.

 

* Creating a share - dito.

 

* Tried to copy a big file over the network from my Windows 8 machine. After 1-2 minutes TotalCommander comes back with "... remove write protection ...". The files that don't work are over 10GB. Copying a small file did work once - but with a second small file I did receive that same error.

 

* I just tried to add a new user. This didn't work. Here's the result. "Harald" is the username. This username was automatically created on a new Windows 8 machine. Windows 8 standard installation procedure creates a username that matches the Windows Live username and I need this username on my unRAID machine too.

 

Mar 12 02:24:41 Tower2 emhttp: shcmd (181): useradd -g users -d / -s /bin/false -c '' Harald |& logger
Mar 12 02:24:41 Tower2 logger: useradd: invalid user name 'Harald'
Mar 12 02:24:41 Tower2 emhttp: shcmd (182): chpasswd <<< Harald:"*****"
Mar 12 02:24:41 Tower2 emhttp: _shcmd: shcmd (182): exit status: 1
Mar 12 02:24:41 Tower2 emhttp: shcmd (183): smbpasswd -L -s -a Harald <<< "*****"$'\n'"*****"
Mar 12 02:24:41 Tower2 emhttp: _shcmd: shcmd (183): exit status: 1
Mar 12 02:24:41 Tower2 emhttp: shcmd (184): cp /etc/passwd /etc/shadow /etc/samba/private/smbpasswd /boot/config
Mar 12 02:24:41 Tower2 emhttp: shcmd (185): :>/etc/samba/smb-shares.conf
Mar 12 02:24:41 Tower2 emhttp: Restart SMB...
Mar 12 02:24:41 Tower2 emhttp: shcmd (186): killall -HUP smbd
Mar 12 02:24:41 Tower2 emhttp: shcmd (187): ps axc | grep -q rpc.mountd
Mar 12 02:24:41 Tower2 emhttp: _shcmd: shcmd (187): exit status: 1
Mar 12 02:24:41 Tower2 emhttp: shcmd (188): /usr/local/sbin/emhttp_event svcs_restarted
Mar 12 02:24:41 Tower2 emhttp_event: svcs_restarted

 

 

So my question is: Can I go back to 4.7 without any problems? Whatever I do with 5.0 opens a bucket of problems for me (seems that it's just me). 4.7 worked for me since it's release...

 

Thanks in advance

Harald

 

[jbartlett]

Please post your entire syslog.

[hawihoney]

I did reboot in the meantime, fresh syslog currently.

 

Just some seconds ago I started the same steps that lead to my post above. If things go wrong again I will post both syslogs (two machines, currently the syslogs are stilll empty, just filled with Telnet Connect notes).

 

What I found out in the meantime:

 

* unRAID does not except mixed characters in a user name. "Harald" is not allowed --> "harald" works. This is bad because this is my name in my Windows environment and I can't change that without a complete re-work of my non-unRAID environment.

 

* What I changed an hour ago is "Workgroup" --> "WORKGROUP". Don't know if this is important. My Windows 8 shows "WORKGROUP" in the settings. unRAID did show "Workgroup" in the settings. I started two copy jobs (2xImgBurn, 28GB and 21GB) to both machines in parallel. Looks good so far.

 

* The Shares: Hmm, I can't change anything. If this happens no line pops up in the syslog.

 

Thanks

Harald

 

[hawihoney]

It's happening again. Copying files from Windows 8 to unRAID does not work here. Nothing, really nothing is added to syslog. Windows simply says "Unknown network error" (Explorer) or "... remove write protection ..." (TotalCommander). But now comes the funny part:

 

The files in question are ISO files (> 15GB).

 

If I copy a complete ISO file with Explorer or TotalCommander it fails. Creating the ISO file with ImgBurn works. This is really stupid. What's the difference between copying a huge file and creating a huge file?

 

<Head shaking>

Harald

 

[JonathanM]

It's happening again. Copying files from Windows 8 to unRAID does not work here. Nothing, really nothing is added to syslog. Windows simply says "Unknown network error" (Explorer) or "... remove write protection ..." (TotalCommander). But now comes the funny part:

 

The files in question are ISO files (> 15GB).

 

If I copy a complete ISO file with Explorer or TotalCommander it fails. Creating the ISO file with ImgBurn works. This is really stupid. What's the difference between copying a huge file and creating a huge file?

 

<Head shaking>

Harald

Do you have a cache drive installed and configured for this user share?

[jbartlett]

The Windows name isn't an issue. My account is "John" on my PC and my NAS user is "john", it authenticates without issue. I don't think the workgroup case is either.

[hawihoney]

Do you have a cache drive installed and configured for this user share?

 

No cache drive. I would like to but this older original LimeTech MD-1500LL has no external SATA port. The 15 internal slots are full. There's plenty of room now that I swapped the old drives and added 3TBs.

 

I've seen lots of funny things but I can't remember something like that. Perhaps TC and Explorer try to reserve 28GB at once for the ISO while ImgBurn just appends the data until the 28GB are copied. This is simply a guess.

 

[hawihoney]

The Windows name isn't an issue. My account is "John" on my PC and my NAS user is "john", it authenticates without issue. I don't think the workgroup case is either.

 

Yes thought so, now that the same error pops up again.

 

[JonathanM]

Do you have a cache drive installed and configured for this user share?

 

No cache drive. I would like to but this older original LimeTech MD-1500LL has no external SATA port. The 15 internal slots are full. There's plenty of room now that I swapped the old drives and added 3TBs.

 

I've seen lots of funny things but I can't remember something like that. Perhaps TC and Explorer try to reserve 28GB at once for the ISO while ImgBurn just appends the data until the 28GB are copied. This is simply a guess.

I've seen timeouts where windows gives up and says the server isn't available when the disks are fairly full or fragmented. It seems like the server tries to identify each sector that needs to be used before it gives the ok for the copy to start, instead of using the already known free space figure to ok the copy and start the process. I suppose it's safer that way, but it sure leads to some problems when the disks are relatively full.

[hawihoney]

It seems like the server tries to identify each sector that needs to be used before it gives the ok for the copy to start, instead of using the already known free space figure to ok the copy and start the process.

 

No, no, it even happens on an empty 1.5TB disk.

 

What I can tell for sure is that prior the failing copy there are lots of reads on the disk.

 

It's really scary. I can reproduce that on two machines. Both machines have completely different hardware. RC11 is installed on both of them.

 

Copying files does not work. No copy from local attached USB device, no copy from Windows 8 over the network. Lots of reads happen and after some minutes the copy stops with an error.

 

If I create the files it works. I can use ImgBurn or even append (>>) many times.

 

Even mounting the ISO on the Windows machine and creating the ISO over the network works - but copy fails. Isn't that funny?

 

Regards

Harald

[limetech]

Do you have a Win7 PC you can try as the copy source?

 

As for user names, right, user names in linux world are all lower-case.  Samba however does the following:

1. Samba checks the username as it is submitted by the Client, if that fails,

2. Samba checks the usename in all lowercase letters, if that fails,

3. Samba checks the username in all lowercase letters with only the first letter capitalized, if that fails, no logon for you.

 

Check 2 is what let's windows name "Harold" match linux username "harold".

 

To make the rules for usernames more flexible I'd have to integrate PAM, which is not used by slackware.

http://en.wikipedia.org/wiki/Linux_PAM

This is not a trivial undertaking unfortunately.

 

 

[hawihoney]

Thanks for your answer.

 

1.) Thanks for the info about user names.

 

2.) No Windows 7 PC here, sorry. I used this 01/31/2013 Microsoft offer to migrate all my PCs to Windows 8 at once.

 

I have no idea what's going on here.

 

[hawihoney]

Copying big files is still not working here.

 

Today I tried to copy a 44GB file to the array. I see 50,000 (!!!) reads in the Web-GUI before Windows 8 says "Remove write protection. Do you want to overwrite the 0 byte file xyz?". I hit Ok. This leads to another 20,000 reads before that same error pops up.

 

After that I do have a 0 byte file on that drive. I can't delete that file with Windows 8 - same error. I start a console and try to rm that file. This leads to another 55,000 reads before I closed the running Telnet session. Now that I closed that Session that drive is still reading and reading.

 

There's nothing in the syslog. The drive is ok - I can copy small files without any problems. If I open that 44GB ISO on my Windows box and write these 44GB with ImgBurn to the same drive the complete file becomes created.

 

I'm on 5.0RC11. The machine is an original LimeTech MD1510/LL. The drive is a 3TB ST3000DM001 that was precleared with JoeL's tool and has 200GB space left.

 

Thanks

Harald

 

[JonathanM]

The drive is a 3TB ST3000DM001 that was precleared with JoeL's tool and has 200GB space left.

1. Does it work any better if you copy to the drive share instead of the user share?

2. Writing large files to a drive that close to full has always been problematic with reiserfs. I try to keep at least 10% free on my drives to avoid it. You said earlier you were working with an empty 1.5TB drive, now you say you are working with a 94% full drive.

[hawihoney]

Thanks for your answer.

 

There's no difference if copying to a disk share (e.g. disk14) or to a share like ISOs (disk13-14). But the cache seems to be different. I could copy all files thru the cache disk - regardless the size. rsync can copy the files as well from the cache to the target. But MidnightCommander fails too copying an 48GB file from cache to diskxx.

 

I'm a Linux noob, but for me it seems that when copying a 48GB file the file system tries to reserve 48GB at once. That leads to thousands of reads - but no writes. This lasts for some minutes and after that time I do receive an error (timeout?).

 

If I create (not copy) an 48GB file on that same drive (e.g. append data like ImgBurn does) this works always. Windows Explorer, TotalCommander, MidnightCommander use the same failing technic. Rsync on the other side seems to work different. Disk shares and User shares fail, while the Cache share works.

 

As a workaround I now keep a cache disk in both machines (I didn't want to). Without a cache disk, if I copy a big ISO, I open that ISO with DaemonTools and recreate that same ISO with ImgBurn. It's nasty but it works.

 

Regards

Harald

 

 

PS: re. disk size: In the past two weeks I've rebuild my two machines. I copied all data to new drives. The old machines had 750GB, 1TB and 1.5TB drives. The new machines have 1TB, 1.5TB and 3TB drives. The size of the drive doesn't matter. The free space seems to be no problem. I can reproduce that on an empty 3TB disk (or 1.5TB disk) as well. And I can reproduce that on two different machines (MD-1500LL and MD-1510LL) with completely different hardware.

 

[hawihoney]

Last time it happened again I tried to look at the unRAID machine. I'm a Linux noob so I tried some commands with the help of Google - so please forgive me if this completely unrelated:

 

root@Tower2:/mnt# lsof | grep -i Big
smbd      5229       hawi   36u      REG       0,12     116064256 70368744177941 /mnt/user/ISOs/BigFile.iso
shfs      5376       root    5u      REG       9,14     116064256            277 /mnt/disk14/ISOs/BigFile.iso

root@Tower2:/mnt# ps -ef | grep 5229
hawi      5229  1565  0 11:34 ?        00:00:00 /usr/sbin/smbd -D
root      7804  7564  0 11:59 pts/3    00:00:00 grep 5229

root@Tower2:/mnt# ps -ef | grep 5376
root      5376     1  1 Mar20 ?        00:24:18 /usr/local/sbin/shfs /mnt/user -disks 16777214 -o noatime,big_writes,allow_other,use_ino
root      7810  7564  0 11:59 pts/3    00:00:00 grep 5376

 

Doesn't this look like two processes using that same file? Remember, I just copy that file with TotalCommander from my Windows 8 box. There's no more app running. "hawi" is my account on the Windows 8 box. "hawi" is a registered user in my unRAID boxes.

 

This is driving me nuts. With 4.7 I never did experience something like that. Today I could copy one single file after spinning up all drives. The second copy failed again.

 

Actually these towers don't make fun any longer. Currently I think I will replace them with something different. A small Buffalo Terastation in the same gigabit network does not complain. I think I will replace the unRAID machines. I can't go back to 4.7 because of my new 3TB drives and copying files with RC11 does not work always. Seems that I'm the only one with this problems.

 

This will be my last post in this (my) thread - no more posts from here.

 

Regards

Harald

 

[trurl]

root@Tower2:/mnt# lsof | grep -i Big
smbd      5229       hawi   36u      REG       0,12     116064256 70368744177941 /mnt/user/ISOs/BigFile.iso
shfs      5376       root    5u      REG       9,14     116064256            277 /mnt/disk14/ISOs/BigFile.iso

 

Doesn't this look like two processes using that same file? Remember, I just copy that file with TotalCommander from my Windows 8 box. There's no more app running. "hawi" is my account on the Windows 8 box. "hawi" is a registered user in my unRAID boxes.

smbd is the process that is copying from your windows machine

 

shfs is the process that is translating user shares to actual disks

 

In other words, this is completely normal.

[hawihoney]

Argh, thanks - was just an idea.

 

I give up. These two unRAID machines will show up in the trade section after building and testing replacement systems. I'm currently investigating the available options. Synologys DS2413 server and DX1211 (24 disks total) sound quit good.

 

Harald