Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Update: We had to stop DRBD because Qmail was queing too many emails... as soon as I stopped DRBD on the secondary box Qmail was able to process the backlog of email. Last night I changed the protocol to "A" and load average seems steady. I've read that "C" actually performs better; perhaps with beefier machines it does, but "A" seems to work well on a Celeron 600 w/ a 100MBit connection. We'll see. And CentOS automatically runs hdparm. As I tried screwing around with different settings, performance *decreased* so I left it alone. Getting 200MB/24MB with -Tt. CD Chris de Vidal said this while chewing gum: > Wonderful program! > > I set up a 60GB and a 20GB partition and have been using them in production for the past 4 days in concert with Heartbeat. Approximately 100 domains are running on a hot-standby cluster and I'm sleeping better at night :-) > > > I noticed that when I did a dump/restore from another box the load average > went through the roof and even shut down many processes like Apache and MySQL. > > So my question: do our boxes meet the minimum/recommended spec? Both are > identical: > DRBD 0.7.10 > CentOS 3.4 (same as RedHat Enterprise Linux 3) > Celeron 600 > 1GB ECC RAM > 120GB drives (don't recall the DMA/RPM speed but I can find out if necessary) > 100MBit Intel NICs connected by crossover (don't recall which chipset but > I can find out if necessary) > > If they do meet spec, any ideas why load peaked so much? I was writing many small files using dump and restore. I could see how someone could DoS us by repeatedly hammering our logs or uploading thousands of tiny files. > > I do have two DRBD partitions on one spindle, but I used two different sync groups. > > Thanks, > CD > > P.S. I just had a thought: perhaps I need to run hdparm.