5.1 server crashes with KERN_INVALID_ADDRESS doing a Normal backup to a disk file

[francini]

I'm having terrible trouble with my Retrospect Server, 5.1.175 with Driver Update 4.2. It crashes with an EXC_BAD_ACCESS/KERN_INVALID_ADDRESS failure at random times. It only seems to do it when it's running a Normal backup to a disk backup file. The crashes appear to be independent of whether or not it's backing up a remote client or a local disk.

 

If I reset the backup set, the backup runs OK. However, on the next run (when it's "Normal" again), it dies.

 

Hardware: XServe DP 1.33 GHz, 2GB Memory, 4x180 GB drives, XServe RAID w/7 180GB drives

 

A typical crash log is included below.

 

Any thoughts?

 

John Francini

Gigunda Group, Inc

Manchester, NH 03101

 

 

Date/Time: 2003-12-15 12:02:56 -0500

OS Version: 10.2.8 (Build 6R73)

Host: fileserver

 

Command: Retrospect

PID: 11405

 

Exception: EXC_BAD_ACCESS (0x0001)

Codes: KERN_INVALID_ADDRESS (0x0001) at 0x43938264

 

Thread 0 Crashed:

#0 0x001bdfb8 in 0x1bdfb8

#1 0x001be1f0 in 0x1be1f0

#2 0x001be26c in 0x1be26c

#3 0x001bf448 in 0x1bf448

#4 0x00206ad8 in 0x206ad8

#5 0x001d43e4 in 0x1d43e4

#6 0x00348e78 in 0x348e78

#7 0x0034a598 in 0x34a598

#8 0x001fa034 in 0x1fa034

#9 0x001fa230 in 0x1fa230

#10 0x0034aaa8 in 0x34aaa8

#11 0x001d43e4 in 0x1d43e4

#12 0x00270384 in 0x270384

#13 0x001d43e4 in 0x1d43e4

#14 0x00271fdc in 0x271fdc

#15 0x001d43e4 in 0x1d43e4

#16 0x00279804 in 0x279804

#17 0x001fa034 in 0x1fa034

#18 0x001fa230 in 0x1fa230

#19 0x001fa034 in 0x1fa034

#20 0x001fa230 in 0x1fa230

#21 0x002798c8 in 0x2798c8

#22 0x001fa034 in 0x1fa034

#23 0x001fa230 in 0x1fa230

#24 0x001cdb10 in 0x1cdb10

#25 0x001fa034 in 0x1fa034

#26 0x001fa230 in 0x1fa230

#27 0x001cdc08 in 0x1cdc08

#28 0x001cdd64 in 0x1cdd64

 

Thread 1:

#0 0x9000508c in syscall

#1 0x90515d0c in BSD_waitevent

#2 0x905156dc in CarbonSelectThreadFunc

#3 0x90020c28 in _pthread_body

 

Thread 2:

#0 0x9003e9a8 in semaphore_wait_signal_trap

#1 0x9003e7c4 in _pthread_cond_wait

#2 0x9051dbf0 in CarbonOperationThreadFunc

#3 0x90020c28 in _pthread_body

 

PPC Thread State:

srr0: 0x001bdfb8 srr1: 0x0200f030 vrsave: 0x00000000

xer: 0x00000000 lr: 0x001bdff4 ctr: 0x00000000 mq: 0x00000000

r0: 0x4002b264 r1: 0xbfffd0f0 r2: 0x004d9000 r3: 0x4002b218

r4: 0x03d79550 r5: 0x00001ac8 r6: 0x00000000 r7: 0x00000000

r8: 0x00000030 r9: 0xbbf76f6f r10: 0x00000060 r11: 0x00000080

r12: 0x03d68f68 r13: 0x00000000 r14: 0x00000000 r15: 0x00000000

r16: 0x00000000 r17: 0x00000000 r18: 0x00000000 r19: 0x00000000

r20: 0xbffffe50 r21: 0xffffffff r22: 0xbffffc00 r23: 0x00051de0

r24: 0x03c3f6b4 r25: 0x0440c508 r26: 0x00000042 r27: 0x0390d000

r28: 0x03d79560 r29: 0x03d7add0 r30: 0x03d68f68 r31: 0x00001870

 

[natew]

Hi

 

Retrospect 5.1 can't handle any single partition over 1 TB in size. How big is the RAID partition? Which partition is the backup set saved on?

 

Nate

[francini]

The partition being backed up is about 140 GB used of 172 GB (it's one whole 180-GB disk on the XServe itself, not in the XServe RAID array; it's being written to a partition on the array that's 345 GB in size.

 

And as an additional data point -- I see the same crashes on 5.0.238 as well.

 

john

 

 

[natew]

Hi

 

When you installed Retrospect 5.1 did you try clean preference files or import the originals from 5.0? I would try clean prefs to start with. I would also try a backup without compression as a test. A reboot of the machine could help too.

 

Retrospect is definitely going to use more memory on a scond backup because it has to match the files in the backup set to those on disk. The error you are seeing is not an internal Retrospect error so memory management on the system or a flaky ram chip is a possible failure point.

 

Nate

 

 

[francini]

Quote:

---

natew said:

Hi

 

When you installed Retrospect 5.1 did you try clean preference files or import the originals from 5.0? I would try clean prefs to start with. I would also try a backup without compression as a test. A reboot of the machine could help too.

 

Retrospect is definitely going to use more memory on a scond backup because it has to match the files in the backup set to those on disk. The error you are seeing is not an internal Retrospect error so memory management on the system or a flaky ram chip is a possible failure point.

 

Nate

 

 

 

---

 

I've tried it both ways. I imported the original preference files, got crashes, and then wiped the backup account, re-created it, uninstalled and reinstalled Retrospect 5.1.

 

I understand that Retrospect is going to use more memory. Note that this was working fine before when the backup host was a G4 (AGP Graphics). I'll definitely check for flaky RAM, since the whole machine has been getting wonky since I've tried running Retrospect on it. (I think most of its 2 GB of RAM was largely idle prior to my inflicting Retrospect and an XServe RAID on the system.)

 

I bloody well wish Apple would implement ECC memory -- like every other major UNIX server vendor does! Then these sorts of errors would get caught by the error-correction system and reported!

 

Sigh.

 

John