AFP Share Issues

[Joe K]

AFP unstable?

Hello all,

First of all, my apologies for posting as a noob in this forum, but it so far I had no success in the noob forum.

I seem to have cnid related problems, and I am looking for some way to verify that this is really the case (possibly someone here can tell me) and I am also looking for a way to remedy the situation.

I've recently installed FreeNAS 8, (see below for full list of hardware) and at first it worked wonderfully, I had write speeds in the 60 MB/s range. Then, after a week or so, I suddenly had a strange error message, and since then write speed is intermittent (it starts to write quite fast, then stops in the middle, then after some wait it writes the rest) and the share keeps unmounting in the middle of copying files. It's quite frustrating. I would like to ask what I can do to narrow down the problem and remedy the situation.

Hardware involved:

A HP MicroServer, with 4GB of RAM and 4x 3TB HD, FreeNAS-8.0-RELEASE-amd64 on an 8GB USB stick.
A MacBook Pro with Mac OS X 10.6.8.

The strange error message:

"Something wrong with the volume's CNID DB, using temporary CNID DB instead. Check server messages for details. Switching to read-only mode."

This appeared in a message box on the Mac, but it does not appear to be a Mac error message. After this message appeared I could not write to the FreeNAS box anymore, until I rebooted. After reboot, I get the behaviour described above: copying files to the FreeNAS box starts writing fast but stopping in the middle, then resuming more slowly. The share unmounts in the middle of copying.

The log file has stuff like this:

Nov 19 21:09:41 freenas cnid_dbd[16715]: read: Connection reset by peer
Nov 19 21:09:41 freenas cnid_dbd[16715]: error reading message header:
Connection reset by peer

(by the way: how do I set the date in FreeNAS? It is obviously not Nov 19)

I sincerely hope someone can point me in the right direction.

Thank you very much for your help,
Joe

 

[Joe K]

Update

Quick update:

I managed to figure out how to log in via ssh and run "dbd -r".

That gave me a bunch of:
Bad CNID in adouble file of '/mnt/path/to/folder/file.ext'
Writing CNID data for '/mnt/path/to/folder/file.ext' to AppleDouble file

FreeNAS is a bit quicker now, but still nowhere as quick as it was when I started using it a week ago. I has stopped unmounting share in the middle of copying, but I still get the jerky copying, and so on.

I don't know what to try nest. Help?

Thanks,
Joe

 

[Joe K]

Another update: still unresolved after all

Hello all,

I see some of you are reading this, even though nobody seems to have anything they want to say?

Anyway: after some further testing it seems I was wrong. The issue with the AFP share unmounting itself in the middle of a file transfer is still there.

Does anyone have any ideas why an AFP share would unmount while copying files? Any idea which log files to look at? Any ideas at all?

Thank you very much for your help.

Regards,
Joe

 

[kranzel]

Hey Joe K,

I am by no means an expert but I was having the exact same issue. I received the same error and removed the Double folder in the corresponding directory, it didn't help. I then removed all the .double folders. Rebooted my freenas box and mac. Upon first connect, the .doubble folders were rebuilt and the problem went away.

A little later on I started having some more issues and discovered one of my RAM sticks was bad, according to memtest. Replaced the stick and things have been smooth sailing since.

I suspect the .doubble folders were corrupt, but in my case the RAM stick was the root cause.

Just thought I would give you my experience.

 

[rawlinson]

I'm having a few issues as well that seem related with AFP. With AFP enabled, during transfers my Freenas-box just goes AWOL after a while. It disconnects and disappears off the radar! No webgui, no SSH, it doesn't even respond to a ping request and needs a hard reset. This would happen a few times a day. I now disabled AFP and everything seems to be running fine since...

 

[pauldonovan]

I've had the same problem where the box just locks up during AFP usage, particularly Timemachine backups. It happens really frequently, then I fiddle with something and I think it's fixed for weeks, then it starts happening again. I'm currently in one of those crash happy phases at the moment. I really wish I could work out what the issue is.

Paul

 

[ProtoSD]

I think it's probably not AFP related because the Windows people are having similar complaints with CIFS. Very few people seem adventurous enough to want to try adjusting their kmem settings in loader.conf to see if makes a difference. It's really a pretty simple thing to try and the few people that have tried it have told me it helped them. The file is /boot/loader.conf, and the simplest variables to try first are vm.kmem_size, and vm.kmem_size_max, try setting max equal to 1/2 to 2/3 of your RAM, and size equal to 1024M. Only do this if you're using the AMD64 version, otherwise you need to recompile the kernel to make the changes or your system may not boot. You'll need to do this from the console or using ssh, make the '/' filesystem writable so you can edit the file by typing 'mount -uw /'. After you finish editing, reboot. Its a simple thing to try....

 

[pauldonovan]

I've tried no end of tweaking of settings in loader.conf. I've tried settings that worked just fine using ZFSGuru on the same hardware, your settings, no settings. There's absolutely no pattern or logic to it. As I said, it will be fine for literally weeks and then suddenly go wrong. What's more annoying is that when I move the box to a location where I can plug it into a display (the TV downstairs) to watch watch the kernel panic crash messages are, it works fine. I think that's because it's connected wirelessly to the rest of the computers on the network and it's not enough bandwidth to overwhelm it. But when I take it back to a location with no display it's connected via gigabit ethernet, and that's where it will crash.

I try stripping back or altering one setting at a time in loader.conf to get it stable, and then think I have found it, only to have it crash weeks later!

Paul


Paul

 

[pgulden]

Hello,
I'm trying to setup an AFP share for LDAP users to use for Time Machine backups. I have successfully setup the AFP share and it's available as a Time Machine backup disk, but I can't connect to it nor can I simply connect to any AFP shares that I setup...

Here is the freenas (8.0.1 RC1) log:

Sep 2 15:50:05 freenas freenas[1702]: Executing: /bin/pgrep -F /var/run/afpd.pid afpd
Sep 2 15:50:05 freenas afpd[4406]: bind(fd, (struct sockaddr *)&address, address_length) failed: Address already in use
Sep 2 15:50:13 freenas afpd[4410]: AFP3.3 Login by user
Sep 2 15:50:14 freenas afpd[4410]: AFP logout by user
Sep 2 15:50:14 freenas afpd[4410]: dsi_stream_read: len:0, unexpected EOF
Sep 2 15:50:14 freenas afpd[4410]: afp_over_dsi: client logged out, terminating DSI session
Sep 2 15:50:14 freenas afpd[4410]: AFP statistics: 0.59 KB read, 0.44 KB written

Any ideas?

 

[plq]

What does work?

I am about to try version 8 out with AFP.

In your experience, what does work with respect to AFP, and what does not work?

BR,

peter

 

[OneStickShortOfaBundle]

I too have started having a CNID DB since I updated to the latest FreeNAS. I get the message "Something wrong with the volume's CNID DB, using temporary CNID DB instead. Check server messages for details! Switching to Read- only mode."

Any assistance would be welcome but it will have to be in simple terms for me I'm afraid.

 

[OneStickShortOfaBundle]

Since updating to the latest FreeNAS I get "Something wrong with the volume's CNID DB, using temporary CNID DB instead. Check server messages for details! Switching to Read- only mode."

As a result I can read from but not write to my server. I am waiting for someone to post a fix for this.

 

[jgreco]

I'm running 8r7584 on a box with AFP, Lion, and Time Machine successfully (or at least it looks that way). However, I will note that AFP performance is horrid. This might be due to the traffic running thru a router onto a different subnet, but both NFS and CIFS run several times faster, and when I was poking around from Lion, I found that if I went into the Time Machine disk and started poking around in the "bands" directory inside the "sparsebundle" directory, if I did an "ls -l" from Lion it would sit there and list about one item per second. I have no idea what it's doing.

 

[jgreco]

Oh, and it took about two days for Time Machine to dump the ~250GB from the OS X box to the FreeNAS TM volume. Not very impressive either.

 

[psg190]

Oh, and it took about two days for Time Machine to dump the ~250GB from the OS X box to the FreeNAS TM volume. Not very impressive either.

Mine took two days just to dump 128GB. My RX rate has been averaging 10.2M. I'm wondering if there is anything that can be done to speed this up.

 

[jgreco]

In my case, I'll note that the storage platform I'm using is a ~2005-vintage 4 drive storage server with an AMD Opteron 240EE (slowish/low power) and a good server motherboard. It seems like many of the lower end (E350, N36L) systems here outperform it. What are you using?

 

[psg190]

In my case, I'll note that the storage platform I'm using is a ~2005-vintage 4 drive storage server with an AMD Opteron 240EE (slowish/low power) and a good server motherboard. It seems like many of the lower end (E350, N36L) systems here outperform it. What are you using?

I'm running an HP Proliant Microserver with an AMD N36L. It's just plain sucking with these 2 day backups.

 

[andymemo]

Hello,
I'm trying to setup an AFP share for LDAP users to use for Time Machine backups. I have successfully setup the AFP share and it's available as a Time Machine backup disk, but I can't connect to it nor can I simply connect to any AFP shares that I setup...

Here is the freenas (8.0.1 RC1) log:

Sep 2 15:50:05 freenas freenas[1702]: Executing: /bin/pgrep -F /var/run/afpd.pid afpd
Sep 2 15:50:05 freenas afpd[4406]: bind(fd, (struct sockaddr *)&address, address_length) failed: Address already in use
Sep 2 15:50:13 freenas afpd[4410]: AFP3.3 Login by user
Sep 2 15:50:14 freenas afpd[4410]: AFP logout by user
Sep 2 15:50:14 freenas afpd[4410]: dsi_stream_read: len:0, unexpected EOF
Sep 2 15:50:14 freenas afpd[4410]: afp_over_dsi: client logged out, terminating DSI session
Sep 2 15:50:14 freenas afpd[4410]: AFP statistics: 0.59 KB read, 0.44 KB written

Any ideas?

Since updating to RC1 Im also having this same issue with mounting AFP shares. It seems that I can only mount my AFP shares as a guest, when I enter a valid username & password authentication fails. Here is an extract from my log:

Code:

Sep 14 20:30:49 NAS avahi-daemon[4006]: WARNING: No NSS support for mDNS detected, consider installing nss-mdns!
Sep 14 20:30:49 NAS freenas[1851]: Executing: /usr/sbin/service netatalk restart
Sep 14 20:30:49 NAS freenas: No matching processes were found
Sep 14 20:30:49 NAS freenas: No matching processes were found
Sep 14 20:30:49 NAS cnid_metad[4021]: Set syslog logging to level: LOG_NOTE
Sep 14 20:30:49 NAS [4023]: Set syslog logging to level: LOG_NOTE
Sep 14 20:30:49 NAS freenas[1851]: Executing: /bin/pgrep -F /var/run/afpd.pid afpd
Sep 14 20:30:50 NAS afpd[4024]: AFP/TCP started, advertising 192.168.0.1:548 (2.2.0)
Sep 14 20:30:52 NAS afpd[4024]: bind(fd, (struct sockaddr *)&address, address_length) failed: Address already in use
Sep 14 20:31:10 NAS afpd[4037]: in openpam_read_chain(): /etc/pam.d/netatalk(1): invalid facility 'b' (ignored)
Sep 14 20:31:10 NAS afpd[4037]: AFP statistics: 0.51 KB read, 0.38 KB written

Also, I find if I stop and restart the AFP service, I occasionally see an error relating to signal termination.

 

[krakoliezene]

AFP problem after installing 8.01 RC1

Hello, I installed 8.0.1 RC1 lately and it generally works well. Especially Time Machine over AFP works again after upgrading to Lion.
I'm having a problem with AFP filesharing though: if I try to copy or move files from one ZFS dataset to another my client shows an error (OSStatus error -36.) and the console shows:
Sep 15 12:40:44 freenas afpd[6814]: AFP3.3 Login by www
Sep 15 12:40:44 freenas afpd[6814]: afp_disconnect: trying primary reconnect
Sep 15 12:40:44 freenas afpd[2807]: Reconnect: no child[3607]
Sep 15 12:40:49 freenas afpd[6814]: afp_disconnect: primary reconnect failed
Sep 15 12:40:50 freenas afpd[6814]: ad_openat: cant chdir back, exiting
Sep 15 12:40:50 freenas afpd[6817]: AFP3.3 Login by www
Sep 15 12:40:50 freenas afpd[6817]: afp_disconnect: trying primary reconnect
Sep 15 12:40:50 freenas afpd[2807]: Reconnect: no child[6814]
Sep 15 12:40:55 freenas afpd[6817]: afp_disconnect: primary reconnect failed
Sep 15 12:40:55 freenas afpd[6817]: ad_openat: cant chdir back, exiting
Sep 15 12:40:55 freenas afpd[6821]: AFP3.3 Login by www
Sep 15 12:40:55 freenas afpd[6821]: afp_disconnect: trying primary reconnect
Sep 15 12:40:55 freenas afpd[2807]: Reconnect: no child[6817]

Does anybody have a clue what is causing this behaviour?

 

[jgreco]

Since updating to RC1 Im also having this same issue with mounting AFP shares. It seems that I can only mount my AFP shares as a guest, when I enter a valid username & password authentication fails.

Ah! Ah! I know that one. It's a bug that was fixed in one of the more recent nightly builds. The "invalid facility b" thing is a typo in one of the config files.