Riccardo, if you suspect EAS devices to causing the load you could set SOGoEASDebugEnabled = YES to have a closer look what is going on.
You say "currently I have 48 workers" ... Q is where is the diff to (WOWorkersCount = 60;)? Are the processes dying? Anything in sogo.log? Regarding memory I'd suggest to increase the mem (if available) until you have a stable env and try to identify which process is consuming the memory. You didn't mention how long it takes until your server gets unstable. Does the mem usage increase slowly? (ps aux) Regards, Thomas Am Donnerstag, Juli 02, 2020 11:08 CEST, schrieb "Riccardo Bicelli" ([email protected]) <[email protected]>: Thanks, I re-followed the ActiveSync tuning guide, changed some values, assigned more RAM (8 GB) to sogo VM and sync is running better. The system is good until the memory utilization goes near 100%. From there CPUs (4 at moment) start spiking (stay at 100% for long times), load5 increases (from 0.05 to 3.00). My new settings are /* ActiveSync */ WOWorkersCount = 60; WOWatchDogRequestTimeout = 59; SOGoMaximumPingInterval = 3540; SOGoMaximumSyncInterval = 3540; SOGoInternalSyncInterval = 60; According to this I adjusted nginx/apache timeouts. Currently I have 48 workers, 100% RAM used and and sogo is telling me that there is no child to handle incoming requests! Have I to increase the memory assigned to vm? Is there a way to limit the amount of memory used per child without cmpromising the activesync functionality and mantain the system healthy? What points I am missing? Il giorno mar 30 giu 2020 alle ore 18:38 Thomas Führer <[email protected]> ha scritto: > > Hi Riccardo, > when you have 50 EAS Users using "push" you have to set WOWorkersCount to a > value larger than 50. -> Maybe 60 - 65 is a good start. > Reference: > https://sogo.nu/files/docs/SOGoInstallationGuide.html#_microsoft_enterprise_activesync_tuning > > Another recommendation from my side is to remove SOGoMaximumSyncWindowSize. > This parameter can cause troubles which some EAS devices. > > Regards, > Thomas > > > Am Dienstag, Juni 30, 2020 13:26 CEST, schrieb "Riccardo Bicelli" > ([email protected]) <[email protected]>: > > > Hi All, > I'm encountering some issues with ActiveSync. > Some clients (gmail app, ios app, outlook app) are losing the > synchronization with no apparent reason. > > Sometimes for restore sync I have to reset device with > > sogo-tool manage-eas resetdevice > > or re-create the account on client device, which is very annoying. > > Installation with approx 50 devices. > In this setup I'm using two sogo servers, one for internal network and > one for external. The two instances have some little differences in > configuration: > > The external one has a different LDAP user filter > (memberOf=ExternalMailUsers...) > The external one is behind an nginx reverse proxy > (client->nginx->apache->sogo) > The internal one is configured as usual (client->apache->sogo) > > Both sogo servers have 4 GB of RAM. > > On both SOGo servers ActiveSync settings are: > > WOWorkersCount = 42; > WOWatchDogRequestTimeout = 12; > SOGoMaximumPingInterval = 850; > SOGoMaximumSyncInterval = 850; > SOGoInternalSyncInterval = 30; > SOGoMaximumSyncResponseSize = 10240; > SOGoMaximumSyncWindowSize = 256; > > WOListenQueueSize = 52; > > On nginx reverse proxy the configuration is: > > location /Microsoft-Server-ActiveSync{ > access_log /var/log/nginx/activesync.log; > error_log /var/log/nginx/activesync-error.log; > resolver xxx.xxx.xxx.xxx; > > proxy_connect_timeout 75; > proxy_send_timeout 900; > proxy_read_timeout 900; > > proxy_set_header Host $host; > proxy_set_header X-Real-IP $remote_addr; > proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; > > proxy_pass > http://xxx.xxx.xxx.xxx/SOGo/Microsoft-Server-ActiveSync; > proxy_redirect > http://xxx.xxx.xxx.xxx/SOGo/Microsoft-Server-ActiveSync /; > } > > On both apache servers: > > ProxyPass /Microsoft-Server-ActiveSync \ > http://127.0.0.1:20000/SOGo/Microsoft-Server-ActiveSync \ > retry=60 connectiontimeout=75 timeout=900 > > > On external SOGo servers, which is the most loaded of two, I see lot > of errors like: > > sogod [29139]: [ERROR] <0x0x5587ddd60480[WOWatchDog]> No child > available to handle incoming request! > > and > > [ERROR] <0x0x5587de41fc40[WOHttpTransaction]> client disconnected > during delivery of response for <WORequest[0x0x5587de139360]: > method=POST > uri=/SOGo/Microsoft-Server-ActiveSync?Cmd=FolderSync&User=xxxxxx%40xxxxxxxxxxxx&DeviceId=androidc778xxxxx5&DeviceType=Android > app=SOGo rqKey=Microsoft-Server-ActiveSync rqPath=(null)> (len=63): > the socket was shutdown > > It seems I do not have enough workers available on external SOGo server. > Sometimes it also happens the server goes swappy. > > I tried to follow the documentation on setting worker numbers, sync > response size, listen queue but i feel a bit lost. > > What are the correct guideline of Server capacity sizing regarding > number of clients? > > thanks > Riccardo > -- > [email protected] > https://inverse.ca/sogo/lists > > > > > > -- > [email protected] > https://inverse.ca/sogo/lists -- [email protected] https://inverse.ca/sogo/lists -- [email protected] https://inverse.ca/sogo/lists
