We added about 350 users to our 2000+ user dual server configuration in the last week and were doing pretty well until this afternoon. Suddenly the CPU load graph stopped looking like its normal Donky Kong video game simulation (up and down) and more resembled a 100% highway with a few dips. Declude processes were taking quite a while to clear before finishing, to be replaced by another. I pulled out some multi thousand line tests and it nary made a dent.
Just before bringing our 3rd server into the fold, things quieted down. While I've already ordered 2 new dual processor 1U's, I want to par down (if not eliminate) the variables invovled: 1) If an external DB query slowed things down, delaying each Declude process, would Declude still show high CPU consumption while waiting and would the graph still be pegged? If not, is there any situation external to my server that would? 2) Is it possible for Declude to be consuming CPU cycles while idling for some other reason? 3) If something else is running in the background, eating cyles, does Declude 'look' like its working harder? 4) If a user (or users) all received masses of attached files (say multi megabyte), would this slow things down in the way described? 5) When a new client reports having 30 users, whats the best way to decipher if this is the case? Is there a log analyzer that inventories unique addresses (understanding that 1 user can have many addresses). Thanks! Dan --- [This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)] --- This E-mail came from the Declude.JunkMail mailing list. To unsubscribe, just send an E-mail to [EMAIL PROTECTED], and type "unsubscribe Declude.JunkMail". The archives can be found at http://www.mail-archive.com.
