Re: [SOGo] EAS huge memory consumption
On 2016-06-10 5:57 AM, Cristian Mammoli (c.mamm...@apra.it) wrote: EAS sogod processes will generally consume 16 to 50 MB of RSS. How comes then that 12 GB are not enough for 90 phones If you set to SxVMemLimit to 300, each sogod child process should be allowed to use about 55 MB of memory of its own before the process is restarted. Bursts can occur because the check only occurs after the request has been sent to the client. So if an EAS client syncs a 200 MB chunk of data, memory will grow a lot larger for that process during response preparation. -- Ludovic Marcotte lmarco...@inverse.ca :: +1.514.755.3630 :: http://inverse.ca Inverse inc. :: Leaders behind SOGo (http://sogo.nu), PacketFence (http://packetfence.org) and Fingerbank (http://fingerbank.org) -- users@sogo.nu https://inverse.ca/sogo/lists
Re: [SOGo] EAS huge memory consumption
and sogo.conf: /* ActiveSync */ SOGoMaximumSyncResponseSize = 1024; SOGoMaximumPingInterval = 10; SOGoMaximumSyncInterval = 30; SOGoInternalSyncInterval = 60; These settings are broken. Reread: https://sogo.nu/files/docs/SOGoInstallationGuide.html#_microsoft_enterprise_activesync_tuning Not all our phones are using push, anyway I changed them this way: /* ActiveSync */ SOGoMaximumSyncResponseSize = 1024; //SOGoMaximumPingInterval = 10; //SOGoMaximumSyncInterval = 30; //SOGoInternalSyncInterval = 60; SOGoMaximumPingInterval = 3540; SOGoMaximumSyncInterval = 3540; SOGoInternalSyncInterval = 60; WOWorkersCount = 100; and ProxyPass /Microsoft-Server-ActiveSync \ http://srvsogo02.apra.it:2/SOGo/Microsoft-Server-ActiveSync \ retry=60 connectiontimeout=5 timeout=3600 in Apache Is this expected behaviour? How many resources am I supposed to use to handle 90 phones? If you have 90 phones, you should have at least 90 sogod processes since an EAS connection using push will consume one process. You can even dedicate a sogod instance for EAS (https://sogo.nu/nc/support/faq/article/dedicated-separate-sogo-instance-for-activesync.html) I'm already using a *dedicated server* for EAS EAS sogod processes will generally consume 16 to 50 MB of RSS. How comes then that 12 GB are not enough for 90 phones -- users@sogo.nu https://inverse.ca/sogo/lists
Re: [SOGo] EAS huge memory consumption
users-requ...@sogo.nu wrote on 09.06.2016 18:05:13: > > Is this expected behaviour? How many resources am I supposed to use to > handle 90 phones? This is really something that needs to be optimized. For comparison: an IBM Traveler server that can handle that amount of users needs just 8 GB of RAM and runs fine with two cores (and does much more other work, too). Kai-Uwe Rommel kai-uwe.rom...@ars.de Telefon+Mobil +49 89 32468-120 ARS Eventtipp: Sichern Sie sich Ihr kostenloses Early Bird-Ticket bei Anmeldung zur ARScon'16 bis 10. Juli! ARS Computer und Consulting GmbH, http://www.ars.de Ridlerstrasse 37, 80339 Muenchen, Deutschland ARS - The Art of Software Engineering Software Engineering, Technologieberatung, IBM Software, IBM Lizenzmanagement Services Handelsregister Muenchen, HRB 101829, USt-ID: DE 155 068 909 Geschaeftsfuehrer: Michael Arbesmeier, Kai-Uwe Rommel, Roland Schock, Joachim Gucker -- users@sogo.nu https://inverse.ca/sogo/lists
Re: [SOGo] EAS huge memory consumption
On 2016-06-09 12:05 PM, Cristian Mammoli (c.mamm...@apra.it) wrote: ProxyPass /Microsoft-Server-ActiveSync \ http://srvsogo02.apra.it:2/SOGo/Microsoft-Server-ActiveSync \ retry=60 connectiontimeout=5 timeout=360 and sogo.conf: /* ActiveSync */ SOGoMaximumSyncResponseSize = 1024; SOGoMaximumPingInterval = 10; SOGoMaximumSyncInterval = 30; SOGoInternalSyncInterval = 60; These settings are broken. Reread: https://sogo.nu/files/docs/SOGoInstallationGuide.html#_microsoft_enterprise_activesync_tuning WOWorkersCount = 40; SxVMemLimit = 384; From the tests I did I cannot set WOWorkersCount lower than 40 or I start getting lots of: No child available to handle incoming request! I cannot set lower than 384 either: I already have lots of: terminating app, vMem size limit (384 MB) has been reached (currently 397 MB) That is normal - as what is considered right now, is the VSZ value - which can be quite large, especially on 64-bit systems. I think we should rather consider the RSS value, which can be more easily tuned and understood. The server has 4 cpus and 12 GB of RAM and after only 6-8 hours all the RAM is occupied and the server start swapping until the oom killer terminates all sogod processes. I have another server with 4 GB of RAM that is handling 200 heavy Thunderbird clients just fine. Is this expected behaviour? How many resources am I supposed to use to handle 90 phones? If you have 90 phones, you should have at least 90 sogod processes since an EAS connection using push will consume one process. You can even dedicate a sogod instance for EAS (https://sogo.nu/nc/support/faq/article/dedicated-separate-sogo-instance-for-activesync.html) EAS sogod processes will generally consume 16 to 50 MB of RSS. Thanks, -- Ludovic Marcotte lmarco...@inverse.ca :: +1.514.755.3630 :: http://inverse.ca Inverse inc. :: Leaders behind SOGo (http://sogo.nu), PacketFence (http://packetfence.org) and Fingerbank (http://fingerbank.org) -- users@sogo.nu https://inverse.ca/sogo/lists
[SOGo] EAS huge memory consumption
Hi, we have around 90 EAS clients (Android Phones) and a SOGo server dedicated to EAS only. Most of the phones are not configured in Push (sync interval=15 min) but I cannot exclude that some user manually configured it in push Apache is configured this way: ProxyPass /Microsoft-Server-ActiveSync \ http://srvsogo02.apra.it:2/SOGo/Microsoft-Server-ActiveSync \ retry=60 connectiontimeout=5 timeout=360 and sogo.conf: /* ActiveSync */ SOGoMaximumSyncResponseSize = 1024; SOGoMaximumPingInterval = 10; SOGoMaximumSyncInterval = 30; SOGoInternalSyncInterval = 60; WOWorkersCount = 40; SxVMemLimit = 384; From the tests I did I cannot set WOWorkersCount lower than 40 or I start getting lots of: No child available to handle incoming request! I cannot set lower than 384 either: I already have lots of: terminating app, vMem size limit (384 MB) has been reached (currently 397 MB) [root@srvsogo02 ~]# grep -i "vMem size limit" /var/log/sogo/sogo.log |wc -l 42621 The server has 4 cpus and 12 GB of RAM and after only 6-8 hours all the RAM is occupied and the server start swapping until the oom killer terminates all sogod processes. I have another server with 4 GB of RAM that is handling 200 heavy Thunderbird clients just fine. Is this expected behaviour? How many resources am I supposed to use to handle 90 phones? Thanks Cristian -- users@sogo.nu https://inverse.ca/sogo/lists