| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Comments:
<0> oh, ok, got it.
<0> hope it's gonna support one sata hdd, two dvdrw and a floppy and two extra aircoolers
<1> .o
<2> okay if you were going to recommend a server lvl uw320 scsi controller what would it be?
<3> whatever the server board ships with onboard :p
<1> nod
<2> just thinking :)
<1> what's that?
<4> !bugs erp
<2> about replacement hardware for work
<5> ok, this is bugging me.
<2> their primary server is a pos home grown box and the strain is starting to show, looking into a real server this time :)
<5> mod_perl 2 for apache2 broken on my Debian install.
<2> bronaugh then quit looking in the mirror ...
<2> dont use apache2, can't help :(
<5> yeah, it's something really weird with perl.
<5> I don't understand what's going on :(
<5> it's not finding Apache2.pm
<5> ... which is in /usr/lib/perl5/Bundle
<5> but only /usr/lib/perl5 is in the path.
<5> so it's clear what's "wrong" -- except that it works on my other server.
<6> I hate MySQL.
<5> uh huh.
<5> it seems most easterners do.
<6> The PoS won't f'n understand that LOCK TABLE means no other process can acquire the lock.
<5> maybe they have more sense than the westerners.
<5> gcbirzan: that -should- work.
<6> In theory, yes.
<5> yeah.
<5> I haven't explicitly -tested- the locking, so who knows how well it actually works.
<5> one q tho - what ver?
<6> 5.0.15
<7> 96.2 requests/sec - 0.6 MB/second - 6.5 kB/request
<7> 1586 requests currently being processed, 139 idle workers
<7> :-D
<5> ok bloody apache shipped with broken config.
<5> gcbirzan: gah.
<5> SpamapS: nice, what's that?
<6> root@pppoe1-dacia:~# ip ro sh | awk '{print $1}' | sort | uniq -d | wc -l
<6> 4
<6> And that's in 2-3 minutes.
<6> (That counts the number of times two processes managed to lock a table.)
<7> bronaugh: Apache 2.2.0, reverse proxying for our 2 busiest servers
<5> nice.
<5> yeah, I just got new hardware at work... yay!
<7> bronaugh: I'm tempted to try the event mpm, instead of worker .. but so far worker has handled the load without any problems.
<5> SpamapS: yeah, don't know what bugs you'd dig up switching MPMs
<0> gcbirzan: those are really dacia's servers?
<2> i gotta remember to hire spamaps when i win the lotter :)
<7> bronaugh: well event is still marked as experimental.. but only because it uses epoll(), but epoll() has been proven for a long time now.
<7> Redragon^: why's that?
<2> cause i think you probably have an apache trick or 2 i dont know hehe
<7> whoa.. a bitchx bold colon.. its been a long time since I saw that!
<7> Redragon^: hehe.. yeah I've learned a few lately
<7> like.. no matter how good an idea it seems ... MaxClients must never be raised beyond 512
<7> on Apache 1.3, I mean.
<5> SpamapS: hmm.
<7> on this 2.2.0 box I have it at 4000
<7> :-D
<7> and it has hit 2200
<5> that's pretty good...
<2> spamaps no you have to change some hard coded stuff to exceed 512 on 1.3 at least
<8> hrm... is there some convention that determines when one should make their programs return 1 or -1 on error?
<5> usually returns a positive #
<7> Redragon^: you have to change hard coded stuff to exceed 256 ..
<5> or 0 if success
<7> Redragon^: you have to be insane to exceed 512. ;)
<2> i thought insanity was a prereq for the job
<7> that too
<8> yeah, success is almost always signified by a 0
<8> but i've errors reperesented by both 1 and -1 equally i think
<5> yeah. doesn't mattermuch, since test is for 0
<8> s/i've/i've seen
<8> i think they just make it return negative numbers if it usally returns a count of something
<8> that might be it
<1> evening Gaia^
<9> hiya
<1> sup
<9> just a reaction to the antibiotics
<2> hiya gaia
<9> hiyas
<1> hey vik
<1> .v Gaia^
<1> heh
<10> hello there, themaxxz
<10> .o themaxxz
<1> heh
<10> .... that could get a bit dangerous, you realise.
<2> .o viking667
<2> could start a trend
<1> .o
<10> see?
<10> .o
<11> hello
<11> i need some help is anyone her
<1> Viking667, huh what danger?
<10> you actually have to be in Fiona's database before she'll listen to you (hostmask included.)
<2> hmm can someone get a good server with 130+ gigs of scsi space (across a raid stripe) and 4 gig of ram for $3500
<6> Yes.
<2> any places you recommend gcbirzan ?
<6> Well, for that kind of money... You can get anything. :-P
<2> thats not alot of money when your talking server quality equipment
<6> Heck, go for IBM. :-)
<6> Won't get you Xeons for that money, but it's... well, quality.
<2> i checked out ironsystems earlier, 3 73gig scsi, raid controller, dual xeon 2.8, 4gig ecc was about 5000
<1> Redragon^, you need rack mountable?
<2> preferred
<6> Well, the drives are expensive.
<2> we have 1 rack at a colo facility and that doesn't last long using towers :)
<2> they got to much crap stuffed into that rack as it is, it probably looks more like a shelf than a rack right now
<6> Heh.
<6> You should see our stuff.
<2> i think i'll lobby for some new legislation, rule #1 get a good admin, #2 listen to him and do what he tells you to do for new equipement
<2> my "i told you so" monster is overweight cause of all the chewin on powers to be
<3> Redragon^ is sata an option for your use?
<2> if its dependable
<2> this is an "i told you so" server so it has to run great hehe
<2> the box they had someone build before is a problem cause of extremely high iowait
<2> which is mostly do to the scsi controller the best i can tell
<2> load runs 2-6 cause of iowait not cause of cpu usage or memory usage
<2> total 3.2% 0.0% 2.1% 0.0% 0.3% 93.9% 0.0%
<2> cpu00 3.9% 0.0% 2.1% 0.1% 0.5% 92.8% 0.1%
<2> cpu01 2.5% 0.0% 2.1% 0.0% 0.1% 95.0% 0.0%
<3> i only have one ancient scsi raid controller (mylex u2w) and that's a slow pos compared to a 3ware raid1
<2> from what i've read the adaptec 2120 has some iowait issues
<10> tojoe: lol. My Adaptec U160 controller must be ancienne then.
<3> Viking667 that mylex is at least 6 years old by now
<10> I'm not sure of the date of mine, but it can't be a lot newer.
<3> along with those ancient ibm hdd's it makes the box really crawl :p
<1> yay!
<9> :)
<9> do /msg x help ban
<12> Evening Gaia
<12> Evening all
<9> longword boo
<1> evening long
<2> hiya longword
<2> longword you recommend any servers? i'm looking for something around $3500 if possible
<12> Redr: Sounds just enough to get a HP DL380
<2> need 3 scsi (probably 73gig), good raid card, ecc
<12> Redr: You'll get a higher spec Dell PE2850 though
<1> longword, tight to get the dual cpu and mem in though
<5> build, RAID array, bulid!
<2> 4gig ecc
<2> dual if i could but i can put that off if i really gotta
<2> trying to replace this pos they got couple years ago
<5> dual core.
<1> standard dl380g4 is 2300euro
<1> but thats single cpu and 1G of ram
<5> cheaper board.
<5> I'm thinking it might take time to sync up the 280G of the new drive that I just added.
Return to
#linuxhelp
or
Go to some related
logs:
#linuxhelp
root inode is not a directory
#c++
ten_words.mp3
#chatzone
wyldroze
dannieashe
#linuxhelp
#linux
#linux
|
|