Hi to everybody. I'm running 2 mailservers with ISPConfig 3.0.1.6, master-master replicated (master-master replication of mysql and /var/vmail with glusterfs). Yesterday one of the server crashed, without giving me even the possibility to enter in ssh. Luckly my server farm gave me the opportunity to enter with serial, and I saw some segmentation fault for clamd, but I was able to see just that error, because the server was not responding to commands. I'm checking the memory to see if it was an hardware fault. The other server is up and running. It is less powerfull that the other one (just 1 gb of ram). I can see that I've a massive usage of memory... and I'm just scare that probably the other server crashed for the same reason... From the server that is now running, I've this usage of memory: Code: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 2578 clamav 20 0 295m 168m 5596 S 0 17.3 0:16.85 clamd 31677 amavis 20 0 202m 85m 4244 S 0 8.8 0:04.68 amavisd-new 31743 amavis 20 0 200m 83m 4296 S 0 8.6 0:04.34 amavisd-new 2144 amavis 20 0 196m 79m 2736 S 0 8.1 0:01.64 amavisd-new 2204 mysql 20 0 230m 45m 6104 S 0 4.7 1:12.65 mysqld 2272 root 20 0 103m 42m 2664 S 0 4.4 0:06.72 spamd 2576 root 20 0 103m 40m 588 S 0 4.2 0:00.02 spamd 2577 root 20 0 103m 40m 588 S 0 4.2 0:00.04 spamd 9667 www-data 20 0 239m 20m 4800 S 0 2.2 0:00.48 apache2 26938 www-data 20 0 239m 20m 4788 S 0 2.1 0:00.30 apache2 27851 www-data 20 0 238m 20m 4816 S 0 2.1 0:01.06 apache2 9665 www-data 20 0 236m 20m 4316 S 0 2.1 0:00.46 apache2 9664 www-data 20 0 236m 19m 4272 S 0 2.0 0:00.26 apache2 27855 www-data 20 0 236m 19m 4252 S 0 2.0 0:00.26 apache2 27852 www-data 20 0 234m 17m 4072 S 0 1.8 0:00.24 apache2 4339 root 20 0 227m 15m 7696 S 0 1.6 0:01.12 apache2 27854 www-data 20 0 228m 12m 3640 S 0 1.3 0:00.14 apache2 9663 www-data 20 0 227m 10m 2500 S 0 1.1 0:00.06 apache2 31093 www-data 20 0 227m 9476 1444 S 0 1.0 0:00.00 apache2 1825 root 20 0 179m 9236 1232 S 0 0.9 16:11.01 glusterfs 9662 www-data 20 0 155m 7452 456 S 0 0.7 0:00.08 apache2 9661 root 20 0 39092 6892 2740 S 0 0.7 0:00.20 vlogger 3007 root 20 0 73356 6224 1984 S 0 0.6 0:10.83 fail2ban-server 31613 postfix 20 0 103m 6024 4468 S 0 0.6 0:00.16 smtpd 31598 postfix 20 0 103m 5816 4292 S 0 0.6 0:00.12 smtpd 31629 postfix 20 0 103m 5812 4296 S 0 0.6 0:00.12 smtpd 2749 root 20 0 184m 4584 1084 S 0 0.5 24:50.60 glusterfsd 2684 clamav 20 0 42776 3884 2644 S 0 0.4 0:00.74 freshclam 31910 postfix 20 0 56504 3456 2672 S 0 0.3 0:00.02 cleanup 31904 postfix 20 0 52268 3224 2432 S 0 0.3 0:00.06 proxymap 31903 postfix 20 0 52152 3204 2460 S 0 0.3 0:00.02 trivial-rewrite 2969 postfix 20 0 41404 3180 2248 S 0 0.3 0:00.12 tlsmgr 3183 root 20 0 66072 3144 2484 S 0 0.3 0:01.94 sshd 31253 root 20 0 66068 3108 2468 S 0 0.3 0:00.32 sshd 31915 postfix 20 0 56408 3044 2332 S 0 0.3 0:00.02 smtp 2871 postfix 20 0 52056 2948 2272 S 0 0.3 0:00.92 qmgr 31751 root 20 0 10828 2416 1052 T 0 0.2 0:00.00 nano 31767 root 20 0 10828 2396 1036 T 0 0.2 0:00.00 nano 2844 root 20 0 36840 2304 1804 S 1 0.2 0:06.36 master 2970 postfix 20 0 38900 2288 1772 S 0 0.2 0:01.00 anvil 31679 postfix 20 0 38900 2272 1788 S 0 0.2 0:00.00 pickup 31999 postfix 20 0 38900 2264 1772 S 0 0.2 0:00.00 showq 3112 vmail 20 0 22920 2060 1316 S 0 0.2 0:18.38 imapd As you can see, clamd is taking 17% of memory, even without usage of CPU. also Amavis is using quite a lot of memory. Is it normal that amavis is using 3 daemons? I saw before that the free memory is jumping from 10MB to 300MB... Is it normal? Michele
Maybe your traffic has increased, that can affect load. Have you seen any spikes in your graphs? Anything strange in log files?
Well, the traffic is increased a lot because before all the traffic was shared between the 2 servers... yesterday one of the servers went down and now all the emails are managed from just one server. Anyway still sound strange to me that clamd is reserving all this memory... thanks Michele
In general, 1 GB of ram is very low for todays hardware. Especiall when you take in account that a GB of ram costs just a few euros. You should increase ram to 2 or 4 GB on that server. You should stop spamd daemon. Spamd is not used by ispconfig as the mail scanning is done with amavisd. Yes. You can set the number of daemons in the amavisd configuration file.
From my own experiences adding more memory will help the increased traffic. Monitor your traffic and try to fine tune your filter and rules.
It was a kind of backup server... the other one, that now is crashed and it's under investigation, has got 2 gb. That is strange, because I was following the perfect server debian how-to... so I don't know then why spamd is up and running. Thanks Michele
In tutorial there is line with following packages Code: apt-get install amavisd-new [B]spamassassin[/B] clamav clamav-daemon zoo unzip bzip2 arj nomarch lzop cabextract apt-listchanges libnet-ldap-perl libauthen-sasl-perl clamav-docs daemon libio-string-perl libio-socket-ssl-perl libnet-ident-perl zip libnet-dns-perl As you see spamassassin is installed and run script starts spamd daemon.
So why it's in the how-to if at the end ispconfig is not using it? Can I even remove it? Anyway the other server is up and running again... and also him have a very strange usage of memory... have a look here: Than few minutes later, the memory free was jumping to 1626024k free... Thanks Michele
I'm not sure if it safe to remove it, but to stop it should be safe. I let Till or someone else answer if it's safe to remove it.
If you would remove it then you would remove amavisd with it and if it would not be needed then it would not have been part of the perfect setup tutorial. So what I told you above is to stop it and not to remove it.
So Till you mean just kill the PID related with spamd? Any idea why the free memory is jumping too much? Could it be the cause of the crash of yesterday in one of the server? (they was online working perfectly for like 8 months) Thanks Michele
ah ok so I'll use Code: /etc/init.d/spamassassin stop Let's see then if I've benefit on the RAM usage... Thanks Michele
Hello, unfortunately also without the spamd process, the usage of memory is jumping, in both servers, really a lot. For example srv1 (4GB ram) is jumping from 1.8GB free to 30MB free... srv2 (1GB ram) is jumping from 500MB free to 20MB free... Any suggestion? Thanks Michele
Thats ok. The Linux kernel uses all free ram to cache harddisk requests. As long as your swap dont get filles 100%, then this is the normal and intended behaviour of the Linux kernel.
Hi Till, thanks for your answer. I'm really happy to know that it's a normal behaviuor... Anyway I still don't know how was possible the crash of srv1 2 days ago... it was something related with the segmentation fault... I've run a test of the RAM, but it was fine, so it was not an hardware fault...
yes, I saved all the /var/log folder... I'm having a look on it, but I can not find any error on it...
