Good morning,
I'm in the proccess of setting up new Cyrus servers. As we are running 3.0.X, versions we have installed 3.0.15. But we are facing a problem. We have copied all the data with sync_client in user mode to new servers. The new servers, are just instances of Cyrus running as like Fastmails calls teraslots (we have created a new automated managing system for this slots and it's in proccess of deciding whether we open it's source). The problem is that when we startup a slot (so a Cyrus proccess), it seems to become flooded of recoveries like the following :
Jul 5 12:10:17 contenedormx-002 imap[90701]: login: proxyserv.server.net [1.2.3.4] test1@xxxxxxxxxxxxx plaintext User logged in SESSIONID=<slot-002-001.proxyserv.server.net-90701-1625479801-1-8353515761537691283>
Jul 5 12:10:26 contenedormx-002 imap[90701]: client id sessionid=<slot-002-001.proxyserv.server.net-90701-1625479801-1-8353515761537691283>: "name" "Mac OS X Mail" "version" "14.0 (3654.100.0.2.22)" "os" "Mac OS X" "os-version" "11.4 (20F71)" "vendor" "Apple Inc."
Jul 5 12:10:41 contenedormx-002 imap[90701]: skiplist: recovered /saremail_slots/slot-002-001/bbdd-funcionamiento-cyrus/domain/s/sonotreal.net/user/n/test1.conversations (3502 records, 382648 bytes) in 1 second
Jul 5 12:10:50 contenedormx-002 imap[90701]: skiplist: checkpointed /saremail_slots/slot-002-001/bbdd-funcionamiento-cyrus/domain/s/sonotreal.net/user/n/test1.conversations (3502 records, 367380 bytes) in 9.344 sec
Jul 5 12:11:25 contenedormx-002 imap[90701]: mailbox: longlock sonotreal.net!user.test1.Enviados for 15.4 seconds
Jul 5 12:11:25 contenedormx-002 imap[90701]: skiplist: longlock /saremail_slots/slot-002-001/bbdd-funcionamiento-cyrus/domain/s/sonotreal.net/user/n/test1.conversations for 15.4 seconds
Until server loads due to this recoveries the server is really not useful because it's extremely slow or non responsibe.....
This Cyrus process has 8000 accounts, some big amount of them with 50GB accounts (and several of them almost full).
I'm suspecting after each restart, as skiplist_always_checkpoint is on by default (commit b5581189219812cc1ed15873368fd50ce7cad2ac) that is causing concurrency and slow checkpoints in the server start. It seems it does checkpoints every 30 minutes anyway (checkpoint event in cyrus.conf). I know it's something about 10 years ago... but I can't figure what else could be... has anyone ever suffered something as this?. We are running Cyrus in FreeBSD as non root user from the beggining (listening in 1430 or 1100 ports) and with the ZFS filesystem... anyone ever has issues with something like that?.
Any help will be very appreciated.
Best regards, -- Egoitz Aurrekoetxea Dpto. de sistemas 944 209 470
Parque Tecnológico. Edificio 103
48170 Zamudio (Bizkaia)
Antes de imprimir este correo electrónico piense si es necesario hacerlo. |
| | |
| 