What kind of chooser are you running? Old versions of LPRng would run the chooser script/program once for every job in the queue, every time it checked. This meant the longer the queue, the longer the load, the longer the time to run the chooser, the more the queue growing and so on......
Running latest LPRng with default chooser settings? On Fri, 12 Mar 2004, Toby Blake wrote: TB> Hmmm, going back to this problem we see when print queues get large TB> and the print server load is extremely high, can anyone explain why TB> this happens (for technical info see previous messages in this thread TB> - I won't repost it all here). TB> TB> We run student labs and times like today, where assignments are due TB> in, result in printers running extremely slowly, but the delay is TB> between the jobs (i.e. the printer is doing nothing between each TB> printed job, the delay is at the server side) and is definitely TB> related to the size of the queue, e.g. TB> TB> [kant]toby: lpq -s -Pat3 TB> [EMAIL PROTECTED] 49 jobs TB> [kant]toby: lpq -s -Pat8 TB> [EMAIL PROTECTED] 36 jobs TB> [kant]toby: TB> TB> [dickinson]lp: uptime TB> 4:03pm up 119 days, 22:49, 1 user, load average: 9.64, 9.18, 7.20 TB> TB> The delay between jobs right now, with this load and these queues is TB> 14 minutes. This delay occurs here: TB> TB> finished '[EMAIL PROTECTED]', status 'JSUCC' at 2004-03-12-15:42:49.171 ## [EMAIL PROTECTED] number=712 process=16602 TB> waiting for subserver to exit at 2004-03-12-15:56:16.463 ## A=<NULL> number=0 process=14349 TB> subserver pid 16602 exit status 'JSUCC' at 2004-03-12-15:56:16.463 ## A=<NULL> number=0 process=14349 TB> TB> ... so after the job has successfully printed, it waits for an lpd TB> process to exit, which takes 14 minutes. TB> TB> What are the lpd processes doing in the time that chews up so much CPU TB> - this is very much related to the size of the queue - if I removed TB> half the jobs in the queue, the delay between jobs would significantly TB> decrease (I know this because I've tried it). TB> TB> Note that our print servers are currently running RH7.1 - I'll be TB> upgrading them to RH9 soon, although I'm not sure this will have any TB> effect. They're running either LPRng 3.8.10 or 3.8.21 - this happens TB> with both. TB> TB> I'd appreciate any help or advice as to what we can do to stop this TB> happening - note again that it only happens when there are a lot of TB> print jobs in the queue, so 99.9% of the time we have no problems TB> whatsoever - I have watched the load increase and the delay between TB> jobs increase as more are added to the queue. TB> TB> Many thanks in advance TB> Toby Blake TB> University of Edinburgh TB> TB> TB> ----------------------------------------------------------------------------- TB> YOU MUST BE A LIST MEMBER IN ORDER TO POST TO THE LPRNG MAILING LIST TB> The address you post from MUST be your subscription address TB> TB> If you need help, send email to [EMAIL PROTECTED] (or lprng-requests TB> or lprng-digest-requests) with the word 'help' in the body. For the impatient, TB> to subscribe to a list with name LIST, send mail to [EMAIL PROTECTED] TB> with: | example: TB> subscribe LIST <mailaddr> | subscribe lprng-digest [EMAIL PROTECTED] TB> unsubscribe LIST <mailaddr> | unsubscribe lprng [EMAIL PROTECTED] TB> TB> If you have major problems, send email to [EMAIL PROTECTED] with the word TB> LPRNGLIST in the SUBJECT line. TB> ----------------------------------------------------------------------------- TB>
smime.p7s
Description: S/MIME Cryptographic Signature
