BAD BAD Memory Chip...

[Wayne 962]

Ten crashes in one day - finally saw the "memory parity error" on the server. Yanked one chip - wrong chip. Yanked the other one - it's been stable now for the past 4 hrs or so.

I've updated and setup into real time the backup server (using MySQL replication) and also replication of the attachments directories. Bottomline, if the server begins to seriously tank, then I can switch over to the back up one in a matter of minutes.

Sorry about the downtime.

-Wayne

[Jack Olsen]

Looks like it's working, now.

[k9handler]

thanks for your hard work keeping it all going Wayne.

[Wayne 962]

I drove to the shop tonite, rebooted the machine, then started the regen on the tables. Then I drove home. Walking up to my house, the cell phone rang stating it had gone down again. Needless to say, I kicked the fence pretty darn hard on my way back out to the car to drive here again.

-Wayne

[beepbeep]

What database? MySQL?

Personally, I don't like MySQL beacuse of absence of stored procedures and problems with rolling back transaction log.

I've deployed dozens of Microsoft SQL servers and never had a single problem with database. Of course, if your main problem is hardware memory fault then software won't do a thing.

My way of solving those problems is buying registered ECC memory DIMM's and downclocking the FSB.

What server are you running database on, how much memory and which sort? If all DIMM-slots are populated you are often forced to back off with agressive memory timing.

[dickster]

wayne,

i am sure you are aware of it but we use "radmin" software for most of our maintenance and its carried out offsite unless there is anything physical to do.

it could save you a few trips(and fence panels)!

[beepbeep]

P.S.

there is really nice utility called "MEMTEST". You boot it off a floppy and it taxes the memory I/O and CPU 100% with different tests until it either passes or find a fault. It takes roughly 1.5 hours though depending on CPU speed and memory installed.

I've solved many tricky problems with this utility...

It might be a good idea to set aside 2 hours for this than doing trial-and-error on live database. Take down the server, pop the MEMTEST off the floppy and run it...if it freezes, downgrade memory timings, FSB speed or start pulling out DIMM's untill it passes.

Here's the link:

http://www.memtest86.com/

[jyl]

Wayne, I'm sorry you're having all this trouble. Thank you for putting in the late hours, and hope your memory returns soon. Maybe eat more fish . . .