Re: [gentoo-user] anyone having apache2 memory issues
On Fri, 2006-03-31 at 20:50 -0800, Alan Bailward wrote: Hey all. I recently swapped my apache 1.3 site (personal blog, albums, etc) over to apache 2 (stable) and all seemed to go well. However lately I've been having a lot of out of memory issues on the server. Even when memory is still available snip What output do you get from: top -b -n 1 For example, here is my apache output: [EMAIL PROTECTED] top -b -n 1 top - 11:48:06 up 2 days, 15:11, 2 users, load average: 0.00, 0.03, 0.04 Tasks: 88 total, 2 running, 86 sleeping, 0 stopped, 0 zombie Cpu(s): 7.4% us, 5.0% sy, 0.0% ni, 86.2% id, 0.8% wa, 0.0% hi, 0.5% si Mem: 2010484k total, 1678576k used, 331908k free, 264884k buffers Swap: 522104k total, 136k used, 521968k free, 945556k cached PID USER PR NI VIRT RES SHR S %CPU %MEMTIME+ COMMAND 15536 root 16 0 56784 10m 4864 S 0.0 0.6 0:00.16 apache2 16374 apache16 0 57188 9816 3108 S 0.0 0.5 0:00.10 apache2 6481 apache15 0 57172 9780 3100 S 0.0 0.5 0:00.02 apache2 Go through top and see what is sucking up your memory. Or you can use gnome-system-monitor, which I like better. Here is a little program I tossed together to free that cached memory. Just run it and specify an amount of memory in MB. Do not run this as root, a normal user is fine. I have 2 GB and after running VMWare I run: mem 1200 Which frees all that cached memory that VMWare sucked up. If you have 1 GB, you should turn swap off before you run this, otherwise you might just be allocating memory from swap and defeat the purpose. Before you run the program, look at the output of free to see how much you should try to free. For example, [EMAIL PROTECTED] free -m total used free shared buffers cached Mem: 1963 1643 3190 258 923 -/+ buffers/cache: 461 1502 Swap: 509 0 509 I have 923 MB sitting in cache. So if I run: [EMAIL PROTECTED] mem 1200 allocating: 1200MB bytes of memory allocated: 1258291200MB bytes of memory freed 1258291200MB bytes of memory I now have 1.2 GB free: [EMAIL PROTECTED] free -m total used free shared buffers cached Mem: 1963 715 1248 0 45 307 -/+ buffers/cache: 362 1600 Swap:509 0 509 To compile the program, just do: You can replace $CFLAGS with whatever you like, for example -O3. [EMAIL PROTECTED] gcc $CFLAGS -o mem mem.c [EMAIL PROTECTED] sudo cp mem /usr/bin BEGIN CUT #include stdlib.h #include stdio.h int main(int argc, char* argv[]) { char* ptr = 0; int i= 0; int size = 0; if (argc != 2) { printf(Usage: %s SIZE\nwhere SIZE is the size of memory to allocate in MB\n, argv[0]); return 1; } /* get the size in MB from the command line */ size = atoi(argv[1]); printf(allocating: %iMB bytes of memory\n, size); ptr = (char*)malloc(1024 * 1024 * size); if (ptr) { printf(allocated: %iMB bytes of memory\n, (1024 * 1024 * size)); for (i=0; i(1024 * 1024 * size); i++) { ptr[i]=1; } free(ptr); printf(freed %iMB bytes of memory\n, (1024 * 1024 * size)); } else printf(failed to allocate %iMB bytes of memory\n, (1024 * 1024 * size)); return 0; } END CUT You might want to try: # mem 225 Jim -- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- I'm a geek, but I don't get it. 36-24-36 = -24. What's the significance? =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Florida, USA, Earth, Solar System, Milky Way -- gentoo-user@gentoo.org mailing list
Re: [gentoo-user] anyone having apache2 memory issues
Jim wrote: Here is a little program I tossed together to free that cached memory. Hm. Why do this? Do you actually get any performance benefits after having freed the memory occupied by the cache? In theory, you shouldn't see any benefits, as the system should throw away memory pages occupied by cache stuff, as soon as there are more important requests (like any malloc). Or am I wrong? Alexander Skwar -- Netscape is not a newsreader, and probably never shall be. -- Tom Christiansen -- gentoo-user@gentoo.org mailing list
Re: [gentoo-user] anyone having apache2 memory issues
On Sat, 2006-04-01 at 22:24 +0200, Alexander Skwar wrote: Jim wrote: Here is a little program I tossed together to free that cached memory. Hm. Why do this? Do you actually get any performance benefits after having freed the memory occupied by the cache? In theory, you shouldn't see any benefits, as the system should throw away memory pages occupied by cache stuff, as soon as there are more important requests (like any malloc). Or am I wrong? That is how it should be. However I noticed when I only had 512 MB of memory that most of my memory would be used and I would see a lot of cache. Instead of that cache being freed or used, I would see a lot of swap file usage which really kills performance. I basically don't want to see swap touched unless I actually run out of physical memory. The best thing to do besides have a bunch of memory is to tune your swappiness: http://kerneltrap.org/node/3000 Now that I have 2 GB of memory, I don't worry about it any more. However when I had 512MB it was an issues, especially when trying to run apache, mysql, postfix, courier and a full desktop. Jim -- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- I'm a geek, but I don't get it. 36-24-36 = -24. What's the significance? =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Florida, USA, Earth, Solar System, Milky Way -- gentoo-user@gentoo.org mailing list
[gentoo-user] anyone having apache2 memory issues
Hey all. I recently swapped my apache 1.3 site (personal blog, albums, etc) over to apache 2 (stable) and all seemed to go well. However lately I've been having a lot of out of memory issues on the server. Even when memory is still available naked alan # free -m total used free sharedbufferscached Mem: 505475 29 0 33 219 -/+ buffers/cache:222282 Swap: 525 75450 vmstat: procs ---memory-- ---swap-- -io --system-- cpu r b swpd free buff cache si sobibo incs us sy id wa 0 0 77020 50124 34212 20667200 2 5 103 126 0 1 99 1 (sorry about the wrapping) I'm still getting errors like this in my site's error log: [Fri Mar 31 20:07:08 2006] [error] [client 209.123.8.19] (12)Cannot allocate memory: couldn't create child process: 12: mt-tb.cgi [Fri Mar 31 20:07:08 2006] [error] [client 209.123.8.19] (12)Cannot allocate memory: couldn't spawn child process: /var/www/arcterex.net/htdocs/mt/mt-tb.cgi[Fri Mar 31 20:07:09 2006] [error] [client 209.123.8.19] (12)Cannot allocate memory: couldn't create child process: 12: mt-tb.cgi This happens for most of the pages when I got to my movable site admin pages (all cgi). I'm using pretty much the stock apache2 config files on a 512mb athlon xp2600 server. I've used threads -threads and mpm-prefork in my USE flags, with pretty much the same results :( The server itself is a standard samba/apache/mysql system which seemed to run mostly ok under apache 1.3, so I'm really wondering what the heck is going on. When the server reports the out of memory conditions I can sometimes hit the page a couple of times and then it'll come up. Other times apache will run with 100% cpu and slowly use up memory and swap until I have to (slowly) login and kill it by hand. Anyone seen anything like this or have an idea on how to fix? TIA Alan -- Alan [EMAIL PROTECTED] - http://arcterex.net Backups are for people who don't pray. -- big Mike -- gentoo-user@gentoo.org mailing list