Unfortunately no one can be told what fun_plug is - you have to see it for yourself.
You are not logged in.
Hi all -
I just recently upgraded my DNS-323 (hardware revision A1) to fw 1.07.
I reboot the device and the LEDs are blue for a while, and everything on it works (I can connect via SMB). After 10 minutes or so, the LEDs turn pink, and samba stops running.
I've rebooted the box using funplug to start a telnet server and done some basic poking. The filesystem is 93% full (33GB free out of 500GB). Samba doesn't seem to keep any logs on the dns-323 (nothing in /var/log) and nothing is in /var/cores. Also, there is a dmesg message for me:
EXT2-fs warning: mounting unchecked fs, running e2fsck is recommended ...some other crap unrelated to filesystems... EXT2-fs error (device md0): ext2_check_page: bad entry in directory #12781014: d irectory entry across blocks - offset=0, inode=3842898016, rec_len=9880, name_le n=23 EXT2-fs error (device md0): ext2_readdir: bad page in #12781014 EXT2-fs error (device md0): ext2_check_page: bad entry in directory #12781014: d irectory entry across blocks - offset=0, inode=3842898016, rec_len=9880, name_le n=23 EXT2-fs error (device md0): ext2_readdir: bad page in #12781014 EXT2-fs error (device md0): ext2_check_page: bad entry in directory #12781014: d irectory entry across blocks - offset=0, inode=3842898016, rec_len=9880, name_le n=23 EXT2-fs error (device md0): ext2_readdir: bad page in #12781014 EXT2-fs error (device md0): ext2_check_page: bad entry in directory #12781014: d irectory entry across blocks - offset=0, inode=3842898016, rec_len=9880, name_le n=23 EXT2-fs error (device md0): ext2_readdir: bad page in #12781014 EXT2-fs error (device md0): ext2_check_page: bad entry in directory #12781014: d irectory entry across blocks - offset=0, inode=3842898016, rec_len=9880, name_le n=23 EXT2-fs error (device md0): ext2_readdir: bad page in #12781014
smbd and nmdb are both not running when I check out the process list with ps (this is after the drive lights go pink). They're both running before the lights turn pink. There's no indication as to why they stop (nothing that looks like a core file anywhere on the system, no logs either).
I noticed that /tmp seems to hold stuff relevant to system status (the chkbutton process touches files in there based upon periodic results of some kernel calls, I guess other processes look for the existence of these files to take some system actions...kinda weird IPC mechanism but whatever works). Here's mine:
# ls /tmp ClientDisplayNonUTF8 max_dl_num re-sch CustomID md0 samba ErrorDisk md0_active scsi_mapping GetTimeServerFinish mount_normal sda LeaseTime msmtp_result.txt sntp_ok QuotaStatus onedisk system_ready boot_finished opserver_frodo temper fan_status pls_do_openssl temper_C fchmod prescan.result temper_F ip_ok prescanbar_result tmp_send.mm ituneprogbar_result raidinfo web_chk makaddr raidup
The most noteworthy is the 'ErrorDisk' file, which I can only assume gets touched because of the ext2 problems that I'm seeing.
I guess I have three questions:
1) where is the pink-light SMB-disabling disk-full limit in disk percentage or GB free?
2) could EXT2 errors be causing my problems with samba dying/pink lights?
3) how do I safely run fsck on my disks?
I'm a tiny bit afraid to fsck on a mounted filesystem with processes running. Does fun_plug get run too late in the boot process to run fsck (e.g. does it get run after all the usual init scripts)? Is there any safe way to run fsck? The fw 1.08 update indicates that I could run fsck from the web interface, has anyone confirmed if this runs safely?
Sorry for all of the question marks. Hopefully some people here have experimented wildly and had some success.
Thanks muchly in advance,
Reid
Offline
reidmefirst wrote:
Is there any safe way to run fsck?
Have you considered fsck from fonz?
http://www.inreto.de/dns323/fsck/
"...you can boot into a small custom Linux to perform manual filesystem checks."
I use this anytime I want to check the disks (and am running raid1 - so need to take the extra step described in the readme)
Have you verified that the problem persists when downgrading the f/w ?
Offline
I did a few things. I downgraded the firmware to 1.06 (which is what I was previously running). It did not fix the problem -- I still got pink LEDs after ~10 minutes, and samba would start but would shut down when the pink LEDs came on. I should mention that the webserver also shut down when the pink LEDs came on.
I ran fsck (I had to do it unsafe...I tried ffp but the fun_plug script for installing it failed). I corrected my filesystem problems. Everything seems to be okay with the filesystem at this point.
My lights still turn pink after about 10 minutes after bootup. Samba still dies about 10 minutes after bootup. The webserver thankfully stays up now. My disks are still 93% full. ErrorDisk file is still created in my /tmp. dmesg shows no filesystem problems now, although it does say that /dev/md0 is an unchecked filesystem and should be checked when it is mounted
I'm kind of at a loss now. I could remove my hard drives and put them in my x86 linux box and run fsck again there. I'd like to know definitively what is causing this 'ErrorDisk' file generation and why my NAS won't love me after a few minutes.
For what it's worth, I run a chroot debian install on my DNS-323. It runs an openssh server, and this has always run fine (even after the pink LEDs light up). So I can sftp files off my NAS
Offline
Is there any new news or solution for this problem?
I've hit the same thing. Overnight, the last two nights, the pink LEDs came on and the samba process was not running, obviously making the file shares inoperative. In the couple of months that I've had the 323, this has also happened once or twice. Rebooting always cleared it up. I can find nothing useful in the logs.
I have fun_plug 0.5 and a few optware packages installed. Overnight, I run a few cron jobs that backup some things on the net to my 323 and I have rsnapshot running. So, there is some intense disk activity every four hours for a few minutes, but that does not involve samba. Further, overnight, there really isn't anything connected to the file shares which means samba shouldn't be under any kind of load.
It sort of looks to me that samba is dying for some reason and some watchdog notices this and changes the light color.
The ramdisk file systems are NOT full. I've read some other threads on this topic and found nothing useful.
I've run fonz fsck to test the drives offline and they are fine. I also have smartctl running and all results from that are OK.
And I am running the "stock" samba that came with the original 323 package.
Should I move to a newer samba, as described in the wiki?
I'm also thinking of looking into tweaking the current samba config to turn on logging. If anyone has some insights into how to do that, please advise.
Thanks!
Offline