#linux

01 November 2005

Total 47 pages. You are browsing page 43/47.

First :: Prev :: [...] [41] [42] [43] [44] [45] [...] :: Next :: Last

22:13 <****> all
22:13 <****> ok
22:13 <****> Thanks for the advice guys; I'm going to join mysql and seek more help from them after I go to the bank
22:13 <****> nemo_work: let me try starting octave and allocating 500+ MB of memory
22:13 <****> bah!
22:14 <****> im gettin ddose on an unused server
22:14 <****> how lame
22:14 <****> Gathond: heh. hm. I could try loading this stupid executable into a perl array
22:14 <****> that might be unhealthy.
22:14 <****> memic: 995 153 842 0 1 38
22:14 <****>
22:14 <****> after allocationg 480 MB of double arrays and quitting
22:15 <****> Gathond: fair 'nuff.
22:15 <****> swap usage is up 100 MB also though
22:15 <****> Gathond: how about my mysql spike theory though?
22:15 <****> Gathond: I guess I have to look elsewhere for why my machine is being so unresponsive...
22:16 <****> Gathond: like, with my oracle example where A x B was close to two gigabytes of ram
22:16 <****> Gathond: a stupid database can very easily try to malloc all physical ram.
22:16 <****> Gathond: even if it releases it later, wiping all traces, surely that could result in unresponsiveness
22:16 <****> nemo_work, mine doesnt respond at all
22:17 <****> nemo_work: mysql could be killed becaus of OOM, but I doubt mysql would allocate that large a chunck
22:17 <****> shoot. why'd I drop by here again
22:17 * nemo_work forgot
22:17 <****> Gathond: dude just the other day in this channel had the OOM killing his mysql. repeatedly.
22:17 <****> Gathond: that was the best I could come up with
22:17 <****> open to other suggestions
22:17 <****> nemo_work: iirc 2.4 kernel takes rather long to spawn processes
22:18 <****> Gathond: I believe he went with hacking the new OOM priority setting
22:18 <****> Gathond: 2.6.x kernel. or else he would have just disabled the OOM
22:18 <****> Gathond: unless you're talking about Ketzu
22:18 <****> so perhaps if mysql recieves alot of reguests that could slow the machines responsiveness to a crawl
22:18 <****> Gathond: and yeah, pthreads *were* slower in 2.4
22:18 <****> but not *that* slow
22:18 <****> a db shouldn't be spawning that many threads...
22:18 <****> nemo_work: he might have had a memory problem:) Im only saying based on Ketzu's free -m I dont think he has one
22:18 <****> and I believe MySQL has a thread pool, like oracle
22:19 <****> nemo_work: one per connection iirc
22:19 <****> Gathond: and while I concede you're right about the buffers, I still think it could be memory spikes
22:19 <****> but anyway. why'd I drop by here :)
22:19 <****> nemo_work: so a busy mysql might potentially spawn alot
22:19 <****> maybe Gathond can help
22:19 <****> how do I make VNCserver serve a currently running desktop?
22:19 <****> there was some kind of swicth for this, IIRC
22:19 <****> Gathond: yeah. well, just like apache there is an initial pool size and a max size
22:19 <****> Gathond: and threads aren't just thrown away
22:20 <****> Gathond: so spawning would be mostly a one time cost 'save for reaping
22:20 <****> Gathond: that in my opinion is even more likely than stupid queries
22:20 <****> Gathond: especially with a decent connection pool lib on the software side
22:20 <****> phogg: x11vnc does that
22:20 <****> nemo_work: true, it keeps threads ready
22:21 <****> nemo_work: the other behaviour is postgres I think, iirc it just has a master process handing each requests to a new process it spawns
22:21 <****> phogg: if you have gnome, the x11vnc stuff is in the gui
22:22 <****> neither mysql not apache nor postgres use threads by default, they just have several processes
22:22 <****> phogg: otherwise, in xorg.conf. add Option "passwordFile" "/root/.vnc/passwd"
22:22 <****> to your Device
22:22 <****> phogg: and load "vnc" to your modules
22:22 <****> nemo_work: using GUIs to configure things is a nice extra but not a necessary feature.
22:23 <****> linuxfreck: in linux 2.4 there was no difference, basically. are you sure mysql does the same on 2.6
22:23 <****> phogg: you will have had to have built Xorg with vnc of course
22:23 <****> Gathond: are you sure that regular vncserver does not? Pretty sure I saw the option somewhere
22:23 <****> Gathond: quite sure
22:23 <****> Gathond: not *much* difference. pthreads do share a little memory :) surely that must be some savings. plus pooling always saves on instantiation
22:23 <****> linuxfreck: but his point still stands, mysql does not spawn a new process at each request, it uses the same process for several requests after one another
22:24 <****> this is also true
22:24 <****> nemo_work: iirc in 2.4 a thread and a process was the same in the kernels view (except threads happening to share memory)
22:24 <****> phogg: quite sure normal vnc does not do it yes
22:24 <****> i have started a second X server wit "xinit xterm -- :2" and X prints out "Xlib: connection to :2.0 refused by server"
22:24 <****> Gathond: on my system in my.cnf, one-thread appears to be a config option, which is disabled
22:25 <****> Gathond: right. sharing memory is always important :)
22:25 <****> ppl how can I send mime attach from cli ?
22:25 <****> nemo_work: yes, but other than that the kernel did not make a distinction about a process and a thread
22:25 <****> yep. we're in total agreement there
22:25 <****> Gathond: nemo_work: wasn't here from the beginning, what's the issue?
22:25 <****> I just don't think mysql is instantiating willnilly
22:26 <****> linuxfreck: Ketzu is claiming performance issues on a machine (probably "LAMP") when serving even static HTML
22:26 <****> nemo_work: it does not, that was an error on my part
22:26 <****> linuxfreck: machine has a mysql db and does dynamic pages
22:26 <****> hi any ideas what could cause /proc/bus/usb/devices not to be there ? although usb devices are reported to be connection when I use dmesg ?
22:26 <****> linuxfreck: Gathond and I were arguing about whether the machine is memory starved
22:26 <****> linuxfreck: I claim dbs can cause huge memory spikes. Gathond thinks it could be pthread instantiation costs in a 2.4.x kernel
22:27 <****> apologies to Gathond if I misrepresent his theory
22:27 <****> nemo_work: I dropped my theory:) but I had it for a minute yes
22:27 <****> I also thought a machine could be memory starved even with much buffer use. I stand corrected there...
22:27 <****> it might be IO starved actually
22:27 <****> I like the db query theory because it is possible to blame stupid programmers ;)
22:27 <****> nemo_work: hehe
22:28 <****> Gathond: btw. wanna know what the query diff was? This was an Oracle 8i db btw
22:28 <****> nemo_work: stupidn programming could cause CPU problems as well I suppose:)
22:28 <****> nemo_work: query diff?
22:28 <****> did I miss something
22:28 <****> Gathond: that query that cost table A times table B memory
22:28 <****> Gathond: in oracle
22:28 <****> ahh
22:28 <****> Gathond: thus several gigs of memory with these tables
22:28 <****> nemo_work: that sounds ugly
22:29 <****> nemo_work: what was it?
22:29 <****> Gathond: select b.a_id,b.foo from b,a where a.id = bar and b.a_id = a.id
22:29 <****> versus
22:29 <****> Gathond: select b.a_id,b.foo from a,b where a.id = bar and b.a_id = a.id
22:29 <****> I kid you not

Total 47 pages. You are browsing page 43/47.

First :: Prev :: [...] [41] [42] [43] [44] [45] [...] :: Next :: Last