Filemaker Backups Running Very Slow

[Adrian M]

We have Filemaker Server 8.0v4 running on a very fast Windows 2003 machine w/ 2gig of ram. I have 2 primary filemaker files being served on this box which are 600mg & 800mg each. Every 3 hours we backup the data using the Filemaker Server backup schedule. When the backup runs the machine bogs down and runs very slow. Even the FM clients run extremely slow in Filemaker. This lasts about 5-6 minutes before normal speed/operation resumes.

Is there anything that we can do about this? I have called Filemaker Support and they didn't have any suggestions. They didn't even provide me with a general idea of how long the backup should take given our file sizes.

Any ideas?

Edited March 8, 2007 by Guest

[Steven H. Blackwell]

First, what is the server cache setting and what is the interval over which the cache is flushed? You can find these in the SAT Tool.

Second, where is the directory to which the backups are being sent? Is it local to the server CPU or is it remote?

Third, what type drives are in this server and what is its clock speed?

Steven

[Adrian M]

The cache size was 64mb but I have increased it to the maximum recommended 457mb and it did not make any differences.

The directory I am backing it up to is the local c drive (default backup location for fm server).

The drive speeds are fine.

[Steven H. Blackwell]

What is the cache flush interval timing? Set it to one minute.

Steven

[xochi]

Steven -- perhaps I'm misunderstanding the cache flush interval, but wouldn't a longer interval give better performance during the flush? E.g. the reason performance is poor is that filemaker is trying to flush the entire cash in one minute?

Or do I have this wrong, and the reason performance is poor during the flush is that the database is essentially paused during the flush, so you want it to be as fast as possible? (But if this were the case, why even allow changes to the cache setting at all)?

hmm...

[Steven H. Blackwell]

The cache flushes continuously. A modern machine should be able to handle 13.3 MB/sec--the theoretical maximum amount.

The built-in backup uses the cache as well.

Backups work like this. Without pausing the files, Server begins to write a backup using its special coding. It skips any busy bits. At the end of the backup, it reconciles the bits it skipped with the new backup.

This is supposed to make backups run faster and with less interference to regular server and network operations.

I suspect that there is something amiss here either with these settings or with the drive sectors whre the backup is being written. But these things are impossible almost to diagnose long distance.

Steven

[LaRetta]

Hi Stephen,

I have the exact same problem and I've been trying to figure it out for a few months. It brings our office to a standstill for 5 minutes. It has always taken 5 minutes no matter the FM version. There are 5 files in /databases but all are only a few kb except our main file which is: design = 1 mb and data = 949 mb. FMS 8.0v4, Win2003, sp1 on Dell 2.80Ghz, 1 GB RAM; backing up to local /backups. Currently, no more than 9 concurrent Users and no web usage. Cache set to 64 mb at 1 minute interval. Not sure I found the drive types - it says they are Dell Perc 4/SC Raid controller and LSI Logic 1020/1030 Ultra320 SCSI and Perc LD0 PRC Raid SCSI Drive device. Is that the drives? But I couldn't find speeds listed like seek per second and wasn't sure I was looking in the right place.

I apologize for jumping it but since our configurations are identical, I decided to. I hope I'm forgiven. I just had a scripted 'write to inventory' not write a single record in the loop. Upon reviewing the logs yesterday, it was fired by a User at the exact moment the backup started. I would give anything to resolve this. ps. Owner says there is nothing wrong with our network so we can't hire someone. I am no network expert (yet, smile). I would sure appreciate anything you could suggest.

LaRetta

Edited March 9, 2007 by Guest

[Adrian M]

No problem about jumping in this thread.

My interval timing is 1 minute.

I do not think it has anything to do with the Filemaker settings because I've experimented with several different cache/interval settings with little effect on performance.

We have a Raid 5 on this system and we are going to update the Raid card drivers this weekend. Also, we will be performing a defrag on the drives to rule that out as the culprit.

When I reboot the server and the Filemaker Service starts up I can look at the Filemaker Server Admin control and the 2 files have a status of "checking" for about 3-5 minutes. This has led me to believe that my files could have some corrupt data or indexes. After exploring into the filemaker files I noticed a few audit tracking fields that are keeping a large amount of data. They haven't been flushed out in over 3 years. During one of my tests I was able to reduce the file sizes by 40% on each file by flushing/archiving.

So...I guess the backup issue isn't necessarily resolved but I have reduced the file sizes to a more managable size and hopefully this will speed up the backups.

I'll post an update next week.

[LaRetta]

Thanks Andrian,

Every so often, I Save As Clone; Compacted. When this is first fired on FM Server, it takes a few minutes 'checking' it; unlike normally it opens instantly. I've been watching this but haven't seen it as a problem (knock on wood). I'd love to archive but this data is too valuable for data modeling and retrieval. I may have to but, FM indicates it can handle much larger files with no problem so ...

Truth? I was getting ready to implement incremental backup process (export of modified records) because I didn't know what else to do. It bothers me to back up posted data (our LineItems are 900,000 records alone) when they will never change. So I thought incremental might be the way to go. But it sounds like 5 minute total-office-standstill is NOT normal so that actually gives me hope to resolve it.

UPDATE: I think we have a RAID which mirrors. Would this effect performance?

L

Edited March 9, 2007 by Guest
Added Update

[mr_vodka]

This thread reminds me of this thread.

[Steven H. Blackwell]

1 GB RAM;

Too low. 2 GB minimum, maybe up to 4 GB is better. Also RAID arrays might be a factor. As I said earlier, these things are virtually impossible to disgnose fully from afar.

Steven

[LaRetta]

I missed that post, John, thank you.

Thank you, Stephen. I shall present your suggestion on the RAM, the RAID and getting a network specialist to review our system. I should have asked before but I've been searching for the answers myself, thinking I just didn't have an FM setting optimized.

The point is ... it is NOT normal behavior and needs to be addressed.

L

Edited March 10, 2007 by Guest

[Adrian M]

OK I have sped up my backup to about 45 seconds.

Solution: We had an audit trail field which logged user information w/ date-time stamps on who last accessed a record. This field was HUGE....so I archived the data & flushed out the audit fields (which were on several tables). This reduced my file sizes by 75%. Pretty incredible. Now the server is running like a champ.

We're still going to address the RAID drivers and defrag the drives to improve the speed some more but I am very relieved we were able to reduce our file size and speed up the backups.

Thank you for all of the recommendations.