Retrospect 4.3 crashes ASIP 6.3.3 Server

[slukes]

OK.. so i've move my backup FROM a OSX Retrospect5 'backup server' due to all the issues it has backing up an ASIP server (errors, server crashing, slow compare)... and now.... its almost worse!

 

 

G4-400 with ASIP 6.3.3 and Retrospect 4.3 backing up 50gb Raid5.

 

Issues are:

 

stability.. more often than not now, the server will 'lock' mid backup, sometimes after 10mins.. 30mins, 1 hour. but it will lock. Memory allocation is 50mb to retrospect, the server has 1.2gb of ram. Retrospect is always loaded. The option to use 'free ram' is off. But still Retrospect crashes.. / locks out the server.

 

Suggestions please on how to make R4.3 and ASIP 6.3.3 stable.

 

Thanks.

 

 

[mcswgn]

Does the log say anything or give any indication of where in the process the freeze is ocurring? Is it only Retrospect that is locking or the whole machine?

 

I doubt that it is memory related as that should give an error, but I would set Retrospect to use free RAM. 50MB is not a lot of RAM for 50GB of data unless there are only a small number of very large files.

 

[mcswgn]

Just to expand on my previous comment--

 

Using the guidelines in the manual, 50MB is enough to handle about 170,000 files. Just to give one data point, which may not match your mix of files at all, I just did a backup of a new client. It has about 6GB on its disk. After matching there were 2.4GB to be written to the backup set. Since this client had not been backed up before, the matched files had to be all from other clients (so a lot of the system files, common applications, etc.). The 2.4GB left amounted to 75,000 files. From my experience, this client is not at all out of the ordinary. If this ratio of files to volume were the same for your 50GB, you would have close to 1,500,000 files and need almost 500MB of RAM to successfully complete a backup.

 

Again, I doubt this is the cause of the freezing, though, as running out of RAM should give an error -108 (out of application memory) and write something to the log instead of freezing.

[slukes]

Hi, I bumped the ram upto 100mb.. still locks and its the whole machine... not just retrpect, i have to reboot the server in the morning.

 

the log shows that is gets through..... x amount of files/time (never the same)

 

 

???

[mcswgn]

Hmm. Maybe things changed from 4.3 to 5.0.238 (I don't remember how 4.3 behaved), but I don't get anything written to the log under 5.0.238 about amount backed up or the time it took until it is all done.

For example, the following is a typical log entry:

 

:4/9/2003 4:23:41 PM: Copying DRIVE D (D:) on UCMS03 (Hodges)?

:4/9/2003 4:23:41 PM: Connected to UCMS03 (Hodges)

:4/9/2003 4:46:17 PM: Comparing DRIVE D (D:) on UCMS03 (Hodges)?

:4/9/2003 4:55:43 PM: Execution completed successfully.

: Completed: 7363 files, 579.6 MB

: Performance: 41.8 MB/minute (31.1 copy, 63.9 compare)

: Duration: 00:32:02 (00:04:21 idle/loading/preparing)

 

The first two lines are written immediately. Then it scans, matches, writes to the backup set, and finally updates the catalog with no additional entries if there are no errors. Then it writes the third line and begins the compare phase, and only when it is all done does it write the last 4 lines. Does the catalog for the backup set show that some files are being written? It would seem like it would almost have to if it is telling you it has completed something. If so, then it has gotten through all of the scanning and matching, which is where it needs all the RAM. Have you tried an immediate backup so that you can watch and see exactly where in the process (scanning, matching, writing, updating catalog, comparing) it gets each time? I realize this may be hard for a production machine that you don't want freezing during the middle of the day.