May developers test policyd with about 2k simultaneous tcp connections? Just open them.
Tobias J. Kreidl ?????: > Hello, Artem: > > On Wed, 31 Oct 2007, Artem Bokhan wrote: > > >> Hello again, I would like to return to the problem. >> >> After creating a lot of connections policyd stops to receive new >> connections and use 100% of cpu. >> Connections was simulated with JMeter (1500 connections). Anybody can >> confirm the same behavior? >> >> > > I don't think we tsted above around 1000 smtp connections, since postfix > is normlly limited in out case to 100 connections on each server. Our > machine has 8 GB of RAM. It was still running, though pretty slowly at > that point. > > Also, you didn't mention of you actually ran out of memory and/or swap > space. Some operating systems, like Solaris, also have a maximum user > process parameter set in the kernel -- maxuprc. Is policyd runnning as > "root" or a different user? You might just write a simple program that > spawns 1500 or so processes and see if you come up against a process > limit, memory limit, or other impediment. If nothing else, it sounds like > you may want to rate-limit the number of connections to postfix or > whatever your mail server is. If that doesn't keep your mail flow going, > you may need additional servers. > > --Tobias > > >> Ubuntu Dapper, uname -a >> Linux 2.6.15-26-amd64-server #1 SMP Thu Aug 3 03:32:26 UTC 2006 x86_64 >> GNU/Linux >> >> >> >> strace: >> >> select(1025, [9 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 >> 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 >> 54 55 56 57 58 59 60 61 62 63 64 65 66 67 69 70 71 72 73 74 75 76 77 78 >> 79 80 81 82 83 84 85 416 417 473 478 482 484], [], NULL, NULL) = 10 () >> select(1025, [9 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 >> 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 >> 54 55 56 57 58 59 60 61 62 63 64 65 66 67 69 70 71 72 73 74 75 76 77 78 >> 79 80 81 82 83 84 85 416 417 473 478 482 484], [], NULL, NULL) = 10 () >> select(1025, [9 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 >> 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 >> 54 55 56 57 58 59 60 61 62 63 64 65 66 67 69 70 71 72 73 74 75 76 77 78 >> 79 80 81 82 83 84 85 416 417 473 478 482 484], [], NULL, NULL) = 10 () >> select(1025, [9 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 >> 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 >> 54 55 56 57 58 59 60 61 62 63 64 65 66 67 69 70 71 72 73 74 75 76 77 78 >> 79 80 81 82 83 84 85 416 417 473 478 482 484], [], NULL, NULL) = 10 () >> select(1025, [9 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 >> 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 >> 54 55 56 57 58 59 60 61 62 63 64 65 66 67 69 70 71 72 73 74 75 76 77 78 >> 79 80 81 82 83 84 85 416 417 473 478 482 484], [], NULL, NULL) = 12 (in >> [81 482]]) >> >> >> sysctl fs.file-max >> fs.file-max = 767080 >> >> >> ulimit -n >> 100000 >> >> >> Compile-time configs: >> >> /* CONFIGS */ >> #define PROJECT "policyd" >> #define VERSION "v1.82" >> >> /* Miscellaneous constants */ >> #define LISTENQ 1023 /* 2nd argument to listen() */ >> #define MAXLINE 1023 /* max text line length */ >> #define BUFFSIZE 8191 /* buffer size for reads and writes */ >> #define BUFSIZE 4095 >> #ifndef MAXFDS >> #define MAXFDS 20479 /* max file descriptors */ >> #endif >> >> >> CC := gcc >> CPPFLAGS := -O $(inc) >> CFLAGS := -g -W -Wall -DMAXFDS=20479 >> OS_NAME := $(shell uname | tr [A-Z] [a-z]) >> LDLIBS = $(lib) -lmysqlclient -lz >> >> >> Cami Sardinha ?????: >> >>> Artem Bokhan wrote: >>> >>> >>>> I will play with ulimit (was 1024 :( ) first and wait until bots will >>>> wake up again :) >>>> May be it's possible to increase the limits by policyd instead of shell?.. >>>> >>>> >>> There is no guarantee that Policyd's attempt to raise it will work. >>> >>> I do recommend you do some kernel tuning as well as raising the >>> ulimit's. I can see from your previous post that Policyd is in fact >>> built to use 4096 file descriptors, however, if you're limiting >>> Policyd to 1024 in ulimit, can you really blame Policyd for its >>> behavior? >>> >>> Cami >>> >>> ------------------------------------------------------------------------- >>> This SF.net email is sponsored by: Splunk Inc. >>> Still grepping through log files to find problems? Stop. >>> Now Search log events and configuration files using AJAX and a browser. >>> Download your FREE copy of Splunk now >> http://get.splunk.com/ >>> _______________________________________________ >>> policyd-users mailing list >>> policyd-users@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/policyd-users >>> >>> >> > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Splunk Inc. > Still grepping through log files to find problems? Stop. > Now Search log events and configuration files using AJAX and a browser. > Download your FREE copy of Splunk now >> http://get.splunk.com/ > _______________________________________________ > policyd-users mailing list > policyd-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/policyd-users > ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/ _______________________________________________ policyd-users mailing list policyd-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/policyd-users
