Problem with WL-HDD and Samba...

[dl9sec]

Hi all,

some days ago i upgraded my WL-HDD from Olegs firmware 1.9.2.7-7c to 1.9.2.7-9. After reconfiguring the whole thing (and after tapping through all the pitfalls concerning the changes between the old and the latest firmware ) all seems to be ok so far.
The NFS mounts are working well and the Samba share appears in my Windows-Workgroup.

But now i discovered a strange problem: I tried to copy a whole directory with some subdirectories an some files in each from my local HDD to a subdirectory on the WL-HDDs share.
The copy process starts, the network LED and the HDD-LEDs were blinking fast. After about 7 seconds the transfer stalls. The both LEDs doesn't flicker anymore instead they were static on. The Samba connection drops (after about a minute the Explorer says, that the drive doesn't exist anymore) and connecting the WL-HDD via SSH was impossible. The WL-HDD seems to hang (not crash, after a minute the HDD-LED starts to flash with a several seconds cycle, but also the webinterface doesn't work anymore).

My WL-HDD starts with the following post-boot script:

#!/bin/sh

# Activate swap 
swapon /dev/discs/disc0/part1

# Set Standby of internal HDD to 10 minutes
/usr/local/sbin/hdparm -S 120 /dev/discs/disc0/disc

# Start dropbear SSH server
/usr/sbin/dropbear

mkdir /tmp/harddisk
# Mount /tmp/harddisk for Samba
mount /dev/discs/disc0/part2 /tmp/harddisk

# wait for /opt to mount 
mount /dev/discs/disc0/part2 /opt 
i=0 
while [ $i -le 30 ] 
do 
  if [ -d /opt/etc ] 
  then 
    break 
  fi 
  sleep 1 
  i=`expr $i + 1` 
done 

# Start Samba and NMBD
/usr/sbin/nmbd -D -l /opt/var/log/nmbd.log -s /opt/etc/smb.conf
/usr/sbin/smbd -D -l /opt/var/log/smbd.log -s /opt/etc/smb.conf

After boot ps shows the following:

  PID  Uid     VmSize Stat Command
    1 admin       580 S   /sbin/init 
    2 admin           SW  [keventd]
    3 admin           SWN [ksoftirqd_CPU0]
    4 admin           SW  [kswapd]
    5 admin           SW  [bdflush]
    6 admin           SW  [kupdated]
    7 admin           SW  [mtdblockd]
   51 admin       380 S   httpd 
   54 admin       408 S   syslogd -m 0 -O /tmp/syslog.log -S -l 7 
   56 admin       364 S   klogd 
   57 admin           SW  [khubd]
   72 1           364 S   [portmap]
   74 admin       504 S   /usr/sbin/statd 
   76 admin           SW  [nfsd]
   78 admin           SW  [lockd]
   79 admin           SW  [rpciod]
   80 admin       484 S   /usr/sbin/mountd 
   83 admin       468 S   watchdog 
   85 admin       344 S   ntp 
   96 admin       480 S   /usr/sbin/dropbear 
   98 admin           SW  [kjournald]
  113 admin       728 S   /usr/sbin/dropbear 
  114 admin       576 S   -sh 
  122 admin       980 S   /usr/sbin/nmbd -D -l /opt/var/log/nmbd.log -s /opt/et
  124 admin       996 S   /usr/sbin/smbd -D -l /opt/var/log/smbd.log -s /opt/et
  126 admin       396 R   ps

The smb.conf is straight forward:

[global]
	workgroup = WORKGROUP
	guest account = nobody
	security = share
	browseable = yes
	guest ok = yes
	guest only = no
	log level = 1
	max log size = 100
	server string = WLHDD-Share

[share]
	path = /tmp/harddisk

[share$]
	path = /tmp/harddisk
	writable = yes
	browseable = yes
	force user = admin

Unfortunately the log smbd.log is empty.
After googling I did already some tries (e.g. with ading oplocks = no, level2 oplocks = no, locking = yes to my smb.conf) but with no success.
With the firmware 1.9.2.7-7c huge filetransfer worked very well.
Now it seems, that the WL-HDD is flodded with data from the network (100Mb/s) and at a specific point HDD transfer is collapsing

Addendum:
I also tried to slow down the Ethernet interface with

et -i eth1 speed 10half

in my post-boot script, but with no success. It seems that the behaviour depends on the amount of data (the copy process stalles at about the same amount of data transceered)...

Any ideas?

Thanks in advance...

Regards, Thorsten

[wengi]

Hi,

mounting is done with fstab since -9 firmware.
Have a look in my howto.

Are you using the intzegrated or the ipkg (Version?) samba?

wengi

[dl9sec]

Hi Wengi,

mounting is done with fstab since -9 firmware.
Have a look in my howto.

The "manual" mount is a relict from the migration from -7c to -9. I have two WL-HDDs and i first modified the one and see if it works before having a non-working NAS...
Will switch over to fstab soon...

Are you using the intzegrated or the ipkg (Version?) samba?

The integrated one...

Thorsten

[dl9sec]

Me again...

further on i tried to follow Wengis translated How-To especially automount with fstab.
Automount of /tmp/harddisk will not work because /tmp/harddisk does not exist anymore in this firmware (therefore i did a mkdir /tmp/harddisk in my post-boot!).
As i follow the instructions exactly (ch. 12a), after a boot there is definitely NO Samba (no smbd/nmbd daemon in process list) working and i can not find any smb.conf.

So i am very confused because the instructions will not work for me. Is there a difference in behaviour of a WL500g and a WL-HDD? Under which circumstances is Samba starting automatically?
Very strange...

Addendum:
Spending the whole evening trying and trying i found out, that -9 behaves a bit different from -7c:
- If Samba is enabled in -7c, /tmp/harddisk is automatically shared via "share" (ro)/"share$" (rw)
- In -9 the Samba that is configured via the webinterface is intended for connected USB devices only. In the moment a USB storage is plugged, the daemons smbd and nmbd are loaded automatically with some kind of default config.

So if someone is starting Samba itself (via rc... or in one of the startup scripts) for sharing the internal HDD the Samba support should be disabled in the webinterface otherwise there will be a conflict and the share doesn't work propperly anymore.

Anyhow, the "startup-sripts/fstab/usb"-stuff is just a side show...the actual problem with the dropping samba at heavy file transfer still exists. The next step will be to install samba2 from the Optware repository and see if it is working better...

Regards, Thorsten

[wengi]

Hmm.. this is strange.

i am using the integrated samba of my WL-HDD 1.9.2.7-9 and it works fine. (I copy a lot of big files)

My setup is as in the HowTo and smb is starting at boot without pre/post entries.
The smb.conf generated by smb:

[admin@blechbuechse root]$ cat /tmp/etc/smb.conf
[global]
        interfaces = br0
        bind interfaces only = yes
        workgroup = WEN
        server string = WLHDD
        guest account = nobody
        security = share
        browseable = yes
        guest ok = yes
        guest only = yes
        log level = 1
        max log size = 100
        encrypt passwords = no
        preserve case = yes
        short preserve case = yes
        client code page = 850

[disc0_3]
        path = /tmp/mnt/disc0_3
        writable = yes
        force user = admin

wengi

[dl9sec]

Hi Wengi,

thanks for the reply.
As announced i just installed xinetd and samba2 as described in your how-to. Installation works fine without any problems. Configuration was too easy via SWAT and what should i say....
...Samba works excellent, perfect, brilliant! I did the same filetransfer as with the integrated Samba daemons running and it was really fast and runs like clockwork.

So i am happy so far and will do all the stuff with my second WL-HDD again

Regards, Thorsten

[dl9sec]

Hi,

me again with some Samba trouble.

After thinking all is ok now with my Samba installation/configuration, i discovered a similar problem like with the "old" Samba.

I have a rar-file (3 parts, with about 95MB/95MB/40MB) on my Samba-share. I started a uncompression via XP file explorer. All started normal. At the WL-HDD i can see the traffic-LED and HDD-LED blinking fast. Exactly after 146.791.575 bytes (139MB) the decompression stalles and the drive dissappears in the file explorer. 7zip reports, that the drive doesn't exist anymore or something with semaphore timeout.

This happens every time exactly at the same time. No more access via SSH is possible anymore, no webinterface acessible. Only the HDD-LED flashes cyclically with some seconds between. The only thing that helps is to unplug from the power source.

The samba log doesn't show anything about that problem. Maybe it has to do with file access. I tried to play with parameters like e.g. "strict sync" but nothing helps.

I think there is a relation between samba, file access and the amount of data that is transferred. There is no problem e.g. streaming GBs of data via NFS with a data rate of several MB/s.

Any ideas?

Regards, Thorsten

[wengi]

Hi,

try to

- copy the files to your pc and extract them loacally
- e2fsck the data partition of the asus
- check the log files
- monitor the 'top' of asus during extraction of the files.

wengi

[dl9sec]

Hi wengi,

thanks for the reply.

- copy the files to your pc and extract them loacally

Done. Works great. The files doesn't seem to be the problem.

- e2fsck the data partition of the asus

Done. No data corruption, HDD is clean (e2fsck -p -f and e2fsck -f).

- check the log files

Done. It is hard to find an entry that has a direct relation with the stall. But around the timestamp of the stall i sometimes can see the folowing entry as an example:

[2009/09/29 21:11:55, 0] source/lib/util_sock.c:read_data(436)
  read_data: read failure for 4. Error = No route to host

and sometimes:

[2009/09/30 19:06:01, 0] source/lib/util_sock.c:read_data(436)
  read_data: read failure for 4. Error = Connection timed out

In the nmbd log i often can find the folowing entry:

[2009/09/30 20:55:48, 0] source/nmbd/nmbd_responserecordsdb.c:find_response_record(235)
  find_response_record: response packet id 16212 received with no matching record.

I don't know if this is "normal" or maybe a hint to the problem i have.

- monitor the 'top' of asus during extraction of the files.

Done. Whenever the WL-HDD freezes, the top output freezes too:

Mem: 29556K used, 616K free, 0K shrd, 1696K buff, 21536K cached
Load average: 1.45, 0.77, 0.31    (State: S=sleeping R=running, W=waiting)

  PID USER     STATUS   RSS  PPID %CPU %MEM COMMAND
  130 admin    R        364   127  1.3  1.2 top
  126 admin    S        392    93  1.1  1.2 dropbear
  131 admin    D       1604   123  0.0  5.3 smbd
  114 admin    S        500     1  0.0  1.6 xinetd
   73 admin    S        476     1  0.0  1.5 statd
  125 admin    S        464     1  0.0  1.5 nmbd
   50 admin    S        356     1  0.0  1.1 httpd
  123 admin    S        356     1  0.0  1.1 smbd

Again and again i forced a check of the HDD with e2fsck. But it says, that all is ok. But i have found the following:

I decompressed the rar. The HDD freezes, i powered off and started the WL-HDD again. Then i renamed the uncompressed fragment and tried to decompress the rar again. And the it works.
For me it seems, that i have something like a bad sector on my HDD and whenever i cross this bad sector, the whole thing freezes. But i don't know how to solve...

Regards, Thorsten

[wengi]

Hmmm... strange
No idea atm.

Does it work if you extract the files directly on the router?

ipkg install unrar
screen
unrar e filename

wengi

[dl9sec]

Hi Wengi,

thanks a million for the hints.

The problem is solved: my HDD is defective. It seems that the BLDC is heating up and something on the controller board of the HDD is then operated outside its temperature range. So i get I/O errors on extensive usage.

I managed to rescue most of my data and put it on a brand new state of the art HDD. One day after rescuing the data the HDD died completely. R.I.P.

So with the new HDD all is working fine and as expected.
All the tips and hints had one good thing. I completely stripped down my Linux on the WL-HDD to the asblolute nescessary and have now a fast and slim OS running with no overhead...

Regards, Thorsten