Jump to content

I can reproduce my "console.log" bug -- ACLS! -- can somebody else?


Recommended Posts

OSX server on Intel xServe -- all patches.

 

When scanning my "data" folder, my console.log file fills up with these:

 

Retrospect(17586,0xa000eca0) malloc: *** vm_allocate(size=1543507968) failed (error code=3)

Retrospect(17586,0xa000eca0) malloc: *** error: can't allocate region

Retrospect(17586,0xa000eca0) malloc: *** set a breakpoint in szone_error to debug

 

So, I looked at *one* of my data folders that only has *one file in it*.

 

Same problem happens.

 

So I removed the ACLS from the folder and rescanned the folder.

 

No errors in console.log!

 

Using WorkGroup manager, I put the ACL back (which was basically just "administrator" with "full control") and save the change -- not even propogating permissions!

 

Rescan the folder.

 

I get this again my console.log file:

 

Retrospect(17586,0xa000eca0) malloc: *** vm_allocate(size=1543507968) failed (error code=3)

Retrospect(17586,0xa000eca0) malloc: *** error: can't allocate region

Retrospect(17586,0xa000eca0) malloc: *** set a breakpoint in szone_error to debug

 

 

Can somebody reproduce this? If so, how do I get the bug reported up?

 

- Steve

Link to comment
Share on other sites

Steve:

 

What version of Mac OS X Server?

What Retrospect version?

 

Could this be related to the earlier ACL Apple bug on Intel Macs for 10.4.7 / 10.4.8, for which there was a workaround (don't back up ACLs, using special hacked version of Retrospect)?

 

It was believed that Apple fixed this bug in 10.4.9, perhaps it regressed to broken somehow in 10.4.10.

 

see:

Retrospect hack for ACL bug on Intel Macs

 

Why don't you try that hacked/patched version using the instructions in the KB article, report results?

 

Russ

Link to comment
Share on other sites

Version of Server: 10.4.10 (all current patches) -- installed the "combo" universal 10.4.10 update when it came out.

 

Version of Retrospect: 6.1.126 with Driver 6.1.11.101

 

I'll try the hacked version of Retrospect and report back!

 

Thanks!

Link to comment
Share on other sites

This appears to have resolved it -- nothing in my console log using this special version of the app!

 

So, who do I report the bug to? Apple? Or EMC?

 

(reported to Apple, but if there's somebody on the forums that can escalate this/reproduce it with EMC, that would be helpful!)

Link to comment
Share on other sites

I can't speak for either Apple or EMC. But it would seem that the right channel is for EMC's developer(s) / support person(s) to contact Apple and file a RADAR report.

 

Realize that the hacked Retrospect version to work around this Apple bug is not preserving your ACL settings, so you would need to reconfigure the ACL metadata on restore.

 

Russ

Link to comment
Share on other sites

We reviewed this post and it sounds like 10.4.10 has introduced a new bug.

 

Engineering would like to get a file system log. Engineering had more details about the API that might be failing, but I don't want to post it in the forum.

 

Can you go into the hidden prefs (option key+preferences) and turn on "file system logging". Reproduce the problem, and email me the operations log? You can use the emcinsignia_forums@emc.com address or my direct email address.

 

Robin

Link to comment
Share on other sites

I would like to add that I am also seeing this error, but in a more random fashion than the thread starter is. I have a fresh install of 10.4.10 on an Intel based xserve with a 1.51 revision apple RAID attached (ACLs on).

 

I am using Retrospect Multi Server 7.5 on a 2003 Server SP1 machine. In my case, the server will show it's status as "Scanning" and will never complete this phase. From my server, the end of my log file:

 

necoCmd: transaction 24: 'DInf' on "retropds"

necoDispatch: transaction 24: result 0

necoCmd: transaction 25: 'Scan' on "retropds"

 

While my Retro. server is stuck on "Scanning", the pitond process on my xserve is at 100% CPU time. My console log from the xserve (client):

 

retropds.24(20975) malloc: *** vm_allocate(size=2348814336) failed (error code=3)

retropds.24(20975) malloc: *** error: can't allocate region

retropds.24(20975) malloc: *** set a breakpoint in szone_error to debug

 

My system volume (no ACLs) will back up correctly every time, my RAID volume did a full backup correctly, and has never completed an incremental backup since. I have defined sub volumes within the RAID, and even the sub volumes will cause the server (or more likely, the client?) to hang randomly.

 

I am working with an EMC Insignia Techincal Support Engineer. If she comes up with a solution I will post back.

 

Thanks - Aaron

Link to comment
Share on other sites

So, I launched the "ACL" version of Retrospect (6.1.126)

 

Turned on the file system logging in the hidden preference

 

Did a "preview" scan of my entire server hard disk.

 

Retrospect quit while scanning.

 

Relaunched retrospect, but the "Operations Log" didn't say anything additional than the normal "Retrospect Version Launched..." information.

 

Nothing at all about the scan.

 

Hidden Preferences --> file system -- file system logging is still checked

 

What am I missing?

 

 

 

I'm going to be on vacation next week, so if you can tell me what I'm doing wrong, I'll be happy to get you what you need today.

 

- Steve

Link to comment
Share on other sites

FWIW... Here's what the log says:

 

 

? $[31]Retrospect$[32] version 6.1.131

launched at 8/17/2007 8:52 AM

+ Retrospect Driver Update, version 6.1.11.101

 

+ Executing $[24]Immediate Backup$[25] at 8/17/2007 9:10 AM

To backup set $[*!s316,,14,+3]STAR BACKUP…

 

- 8/17/2007 9:10:06 AM: Copying $[*!s312,,14,+3]Server HD…

xvolRWHeader: opening header file, current data file "Retrospect Data"

8/17/2007 9:11:39 AM: Execution completed successfully.

Link to comment
Share on other sites

Quote:

We have reproduced this problem. We are investigating the issue, and will report a bug to Apple. It does look like ACL backup fails in 10.4.10 but does work in 10.4.9

 


Robin,

 

You might want to update the KB article to target 10.4.7, 10.4.8, and 10.4.10 so that others can avoid this issue.

Link to comment
Share on other sites

Just to make it clear, Steve M. found a nice bug in the OS. 10.4.10 is not correctly backing up ACL's. In 10.4.8 Retrospect would crash trying to backup ACL's. In 10.4.10 most users don't crash, but an error is displayed in the console.

 

The only way to get a backup of ACLs is to use 10.4.9.

 

We are reporting this to Apple as a bug.

Link to comment
Share on other sites

Quote:

The only way to get a backup of ACLs is to use 10.4.9.

 


Is that correct? My understanding of the ACL bugs is that they only have been shown with the Universal Binary version of Mac OS X. I had understood, at least from my testing and the KB article, that the PPC (non-Universal Binary) Mac OS X, both server and non-server, 10.4.7, 10.4.8, 10.4.9, and 10.4.10, did back up ACLs (mostly - there's a few nits of metadata that may be missed missed, I believe) without crashing and without the log bug that Steve Maser reported.

 

Russ

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...