On 23 Feb 1998, Guy Maor wrote: > We obviously have extreme problems with master's load. My archive > scripts which used to take about 1.5 or 2 hours to run a few months > ago now take 10 hours or more. I think the mailing lists are the main > culprits here. The addition of the real time black hole stuff was the > proverbial straw. Hopefully a new listserver will be christened very > soon.
It isn't the RBL stuff, it only acts on incoming traffic which is minimal (~10msgs/hour). I did some tests today, master without the lists running today (busy day for cron jobs) cruises along at a load average of 5 to 8. Chuck the lists in an the machine kills itself at 30 - 50. Remove the cron jobs and let the machine just handle lists and it is happy at 5-10. For comparision va under the strain of the lists (all out 120 outgoing and about 7msgs/sec incomming - acting as a relay for master) doesn't go past 5. I do not know why your cron jobs are so much slower, I have seen them steadally take longer and longer, today one has been running in excess of 8 hours already. Perhaps there is a bug? Hopefully mark will reboot the machine tomorrow and get the new kernel on, I suspect that initself will give a big performance gain (.32 is much faster than .27 I have noticed) and the tweaks to the drivers should help significantly as well. Oddly, with a load of 28 top reported about 50% cpu utilization indicating gobs of disk bound processes, considering there are 3 disks on two controllers and the load is pretty even across them I think the machine is at it's IO limit. I too am hoping for a new list server, but I haven't heard anything other than rumors. Jason -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] .
