Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hi, I ran into a problem last week when one of our servers went down hard. A lot of information was lost (had to restore from backup). What I think is happening is data is not being synced to disk in a timely manner. I can replicate this with just one server using drbd. Here are my steps to reproduce. a. make partition primary, so /proc/drbd is in state Primary/Unknown b. mount partition c. copy linux src tree to partition d. hard reboot. The write cache on the ide disks are meant to flush every 5 seconds. The write cache in linux is meant to flush every 30 seconds by default I believe or if the cache is 40% full. I have tried waiting at least 30 seconds and still saw the error. I haven't tried doing an actual `sync` command, so I'm not sure if it's the linux buffer, or the ide buffer. I suspect linux, as I did try the same test on a reiserfs partition and didn't lose any files. I am continuing to research this, if anyone has a suggestions please let me know. Cheers, Scott -- -------------------- Gossamer Threads Inc. ---------------------- Scott Beck Email: scott at gossamer-threads.com Lead Software Developer Phone: (604) 687-5804 http://www.gossamer-threads.com Fax: (604) 687-5806