unRAID Server Release 5.0-rc12 Available

[trurl]

Most likely ran out of memory with the 4 preclears. See here

[burtjr]

Try to stop samba and unmount the drives from a telnet session look here (cleanly stop the array from command line) http://lime-technology.com/wiki/index.php/Console#To_cleanly_Stop_the_array_from_the_command_line , reboot and restart new permissions. Wait for all your disks to complete before you do anything else.

[madshi]

Thanks guys. I've tried to unmount the drives via telnet, but it didn't work. The unmount commands froze. After trying for a while I found that there was a defunct "shfs" process with "init" as the parent which had open file handles to some of the drives. I failed to clean this up properly. In the end I had no other choice than to cold reboot. Here's the syslog:

 

http://madshi.net/syslog.txt

 

My telnet repair stuff isn't in there. I first copied the syslog before trying to clean things up. On a quick check I didn't find an "out of memory" log entry. However, there's an "Oops" in the log, plus a few Call Traces. Maybe it's worth looking into.

 

Anyway, after the reboot I've now first applied the new permissions, then ran a checkdisk on all disks (except parity, of course). No problems, everything's fine, for now.

[smakovits]

My mobo is running a Realtek 8111C...

It would be good if someone with direct experience had replied, but ...

 

It is my understanding that all but a small number of older Realtek LAN chipsets are working with the current build.

 

RC12a works fine with my rig and I have the RTL8111C

 

Just curious, before the fix did you have any issues?  There was a point my whole system would simply freeze up and I had to hard reset the thing.  Since adding the Add-in NIC it was fine, but if I can revert back removing the additional NIC, I will.

 

[smakovits]

Summarized: use it, its the best v5 version.

 

... except for stale nfs file handles, in which respect rc10 was far better.

 

So I assume RC12 still has stale file handle issue with NFS?  Or did I misunderstand

[PeterB]

... except for stale nfs file handles, in which respect rc10 was far better.

 

So I assume RC12 still has stale file handle issue with NFS?  Or did I misunderstand

You understand correctly.  rc10 is not absolutely perfect in this respect, but it is far, far, better than rc11/12

[doubleohwhatever]

rc10 is not absolutely perfect in this respect, but it is far, far, better than rc11/12

Agreed. I tried rc11 and then rc12. I just gave up and went back to rc10. rc10 is just about as solid as can be for me.

[madshi]

I'm having some not-so-good experiences with 5.0-rc12a here. I've now had the 2nd time that the web page stopped responding. This time it was definitely not an out of memory problem. Also telnet and name resolution stopped working (error 23) this time. On the other hand, Samba access per IP address still worked. The start of the trouble was that during disk reconstruction one of the disks stopped working (first timeout, then read errors for every sector). The drive not responding is probably not unRAID's fault, but web and telnet stopping to work is very very bad. I've never had this problem with 4.7. The web control always continued to work with 4.7, no matter what else happened. That said, I haven't had a drive failing during risk reconstruction with 4.7, either, so maybe the same problem would have occurred with 4.7, too? Don't know... Here's a syslog:

 

http://madshi.net/syslog2.txt

 

I've shorted the syslog. At the end were millions of read errors...

[RobJ]

I'm having some not-so-good experiences with 5.0-rc12a here. I've now had the 2nd time that the web page stopped responding. This time it was definitely not an out of memory problem. Also telnet and name resolution stopped working (error 23) this time. On the other hand, Samba access per IP address still worked. The start of the trouble was that during disk reconstruction one of the disks stopped working (first timeout, then read errors for every sector). The drive not responding is probably not unRAID's fault, but web and telnet stopping to work is very very bad. I've never had this problem with 4.7. The web control always continued to work with 4.7, no matter what else happened. That said, I haven't had a drive failing during risk reconstruction with 4.7, either, so maybe the same problem would have occurred with 4.7, too? Don't know... Here's a syslog:

 

http://madshi.net/syslog2.txt

 

I've shorted the syslog. At the end were millions of read errors...

 

1. You have no highmem at all, so not surprising that multiple Preclears crashed the system.  You should also be aware that you are at higher risk of the growth of the syslog itself running you out of memory, such as is caused by syslog growth due to large and redundant error message series of lines (like you had).  Please consider adding an additional gigabyte.

 

2. Your 'drive not responding' issue seems very similar to Roadini's here, except for no 'BLK_EH_NOT_HANDLED' flag.  I believe there are also other similar cases elsewhere too, where a drive stops responding to a sas controller, then web access is also lost (no idea why).  The drive and controller stopped communicating, and only 21 seconds later, the system disables the drive, after which you can completely ignore all errors that follow.  Check into a firmware update for your card.  This Firmware upgrades wiki page may help, but I doubt all possible firmware info is on it or linked from it yet.

 

3. Your motherboard SATA ports appear to be in IDE mode, although you currently aren't using them.  Once changed to a native SATA mode such as AHCI (in your BIOS settings), they are probably your fastest and most reliable SATA ports.

[RobJ]

Thanks guys. I've tried to unmount the drives via telnet, but it didn't work. The unmount commands froze. After trying for a while I found that there was a defunct "shfs" process with "init" as the parent which had open file handles to some of the drives. I failed to clean this up properly. In the end I had no other choice than to cold reboot. Here's the syslog:

 

http://madshi.net/syslog.txt

 

My telnet repair stuff isn't in there. I first copied the syslog before trying to clean things up. On a quick check I didn't find an "out of memory" log entry. However, there's an "Oops" in the log, plus a few Call Traces. Maybe it's worth looking into.

 

Anyway, after the reboot I've now first applied the new permissions, then ran a checkdisk on all disks (except parity, of course). No problems, everything's fine, for now.

 

One more quick note, it looks like you attempted (or the system erroneously did because it had crashed!) to restart emhttp.  It cannot be restarted in UnRAID v5.  Apparently some were able to do it, without apparent harm, in v4.7, but it wasn't designed to be restarted, and it definitely does not succeed in v5.

[madshi]

1. You have no highmem at all, so not surprising that multiple Preclears crashed the system.  You should also be aware that you are at higher risk of the growth of the syslog itself running you out of memory, such as is caused by syslog growth due to large and redundant error message series of lines (like you had).  Please consider adding an additional gigabyte.

 

First of all many thanks for analyzing my syslog(s). Is the missing highmem caused by a bad configuration (if so, how can I fix that? I'm a Linux noob)? Or do I simply not have enough hardware RAM in my tower?

 

2. Your 'drive not responding' issue seems very similar to Roadini's here, except for no 'BLK_EH_NOT_HANDLED' flag.  I believe there are also other similar cases elsewhere too, where a drive stops responding to a sas controller, then web access is also lost (no idea why).  The drive and controller stopped communicating, and only 21 seconds later, the system disables the drive, after which you can completely ignore all errors that follow.  Check into a firmware update for your card.  This Firmware upgrades wiki page may help, but I doubt all possible firmware info is on it or linked from it yet.

 

Thanks, I'll check if I can find a firmware update.

 

Is this something Tom should look into? I mean eventually losing web access when a drive stops responding to a sas controller? I think that's a pretty big problem. FWIW, I don't think this happened in 4.7, at least I never experienced it there.

 

3. Your motherboard SATA ports appear to be in IDE mode, although you currently aren't using them.  Once changed to a native SATA mode such as AHCI (in your BIOS settings), they are probably your fastest and most reliable SATA ports.

 

Ok, I guess I should do that then, thanks. I wasn't aware that the motherboard SATA ports can be so "good".

[smakovits]

rc10 is not absolutely perfect in this respect, but it is far, far, better than rc11/12

Agreed. I tried rc11 and then rc12. I just gave up and went back to rc10. rc10 is just about as solid as can be for me.

 

I might be nuts, but to me something like would seem like  show stopper to a full release.  I know many want the full version released and there is now a poll for it, but slow write speeds is one thing but non-functional NFS seems pretty major.

 

I wonder if there is any thought as to why it is having such issues.

[NAS]

rc10 is not absolutely perfect in this respect, but it is far, far, better than rc11/12

Agreed. I tried rc11 and then rc12. I just gave up and went back to rc10. rc10 is just about as solid as can be for me.

 

I might be nuts, but to me something like would seem like  show stopper to a full release.  I know many want the full version released and there is now a poll for it, but slow write speeds is one thing but non-functional NFS seems pretty major.

 

I wonder if there is any thought as to why it is having such issues.

 

Most people use CIFS not NFS so theres less pressure to fix.

[CrashnBrn]

Do we know how close the 5.0 release is? I'm checking daily

[Joe L.]

Do we know how close the 5.0 release is? I'm checking daily

According to this post:

http://lime-technology.com/forum/index.php?topic=22749.0

should be sometime in October 2012.  :'(

[CrashnBrn]

Do we know how close the 5.0 release is? I'm checking daily

According to this post:

http://lime-technology.com/forum/index.php?topic=22749.0

should be sometime in October 2012.  :'(

 

That made me laugh, and then go a sad "Oh" right away.

[umirza85]

So currently I have a 32bit cpu, I was just waiting until 5.0 to move to 64 bit (since im assuming thats when 64bit support will be added). Since this rc is the last one before 5.0 final would it be a good time to upgrade now? or do i still have some time?

[Joe L.]

So currently I have a 32bit cpu, I was just waiting until 5.0 to move to 64 bit (since im assuming thats when 64bit support will be added). Since this rc is the last one before 5.0 final would it be a good time to upgrade now? or do i still have some time?

You have more time.    You are assuming wrong.  If it is added, I'd guess at least 5.1, or later.    It would need a full beta/rc series just of its own.

 

You can upgrade your hardware at any time..., but there is no rush.  (especially the way lime-tech seems to work)

[Dougy]

Since this rc is the last one before 5.0 final would it be a good time to upgrade now? or do i still have some time?

 

I must have miss the announcement that this would be the last RC.

 

If this RC is the last RC before final then that means that the final will be identical to this RC.  If we already know that this RC is going to be released as final then why hasn't it already happened??

 

Like I said I must have missed something

[olympia]

I must have miss the announcement that this would be the last RC.

 

It is in announced in the very first sentence of this thread...

[Dougy]

I must have miss the announcement that this would be the last RC.

 

It is in announced in the very first sentence of this thread...

 

That is not the case.  The announcement is that it is hoped that this will be the last RC.  It will not be known if it will be the last RC until it is tested. 

 

As I said, if we already knew that this was the last RC then why hasn't it already been released as final???

[RobJ]

1. You have no highmem at all...

 

Is the missing highmem caused by a bad configuration (if so, how can I fix that? I'm a Linux noob)? Or do I simply not have enough hardware RAM in my tower?

 

You have 1 gigabyte installed, and it appears that about 128MB of that may be reserved for onboard video.  Lowmem generally takes up about 890MB, so that does not leave anything for highmem.

 

Thanks, I'll check if I can find a firmware update.

 

Is this something Tom should look into? I mean eventually losing web access when a drive stops responding to a sas controller? I think that's a pretty big problem. FWIW, I don't think this happened in 4.7, at least I never experienced it there.

 

Any answer here would be speculative only.  When I look in the syslogs at the output of all of the sas modules, NONE of them look mature to me, with lots of developer statements still appearing.  It's hard for me to consider any of the SAS support, even in the latest releases, to be very mature yet, so I have to lean toward a probable bug in either the sas support modules (in your case, mvsas) and/or the current sas card firmwares.  As JBOD users, we're a small market, much less important than the RAID market, so it's not surprising that JBOD stability is not as high a priority to the card makers.

[dgaschk]

1. You have no highmem at all...

 

Is the missing highmem caused by a bad configuration (if so, how can I fix that? I'm a Linux noob)? Or do I simply not have enough hardware RAM in my tower?

 

You have 1 gigabyte installed, and it appears that about 128MB of that may be reserved for onboard video.  Lowmem generally takes up about 890MB, so that does not leave anything for highmem.

 

Try set set the video RAM to 32MB in BIOS. Choose the lowest setting; the default setting, [automatic], is not correct.

[madshi]

Thanks RobJ and dgaschk. I'll check the BIOS and plan a RAM upgrade for the near future.

 

There isn't a mainboard available with 16 SATA ports, so we can get rid of SAS, or is there? 

[garycase]

Realistically, no ... although you CAN get a board with 14 SATA ports:  http://www.newegg.com/Product/Product.aspx?Item=N82E16813157327

 

I'm not aware of anything with more -- and obviously a board with 8 ports and one add-in 8-port MV8 card is a lot more economical