Sidebar
endurox:v7.5.x:manuals:tpcached.8
Name
tpcached — Enduro/X cache maintenance daemon
DESCRIPTION
tpcached is daemon process started by client process monitor (cpmsrv) (it needs to be registered in <clients> section for Enduro/X application domain). Only one copy of tpcached can be stated per cache database. If no CC tags are used, then one copy shall be run over the app domain. If CC tags are used for cache configuration, then multiple tpcached daemons can be started, per CC tag. The daemon performs following functions:
- Processes any cache databases if expiry flag is present. The daemon tests if records add time (UTC) plus expiry seconds are less than current system UTC time, then record is assumed expired and is deleted.
- Process any cache databases for which clrnosvc flag is present. In this case daemon tests that if cached record’s service name is not advertised in shared memory, then record is deleted from cache.
- Process any limited cache databases having flags lru (last recently used), hits (number of record hits) or fifo (first in, first out) order. The process copies all records to allocated linear array and performs qsort to find records which are beyond the limit. If there are any such records over limit value, then those records are deleted. If duplicates were allowed for database (timesync flag set) then for limits mode, during the scanning, duplicate records are deleted too.
- Process any database for which scandup flag is set. During this mode any duplicate records found are deleted.
In case if cluster operations are configured and flag bcastdel is present, then in case of record removal, this event is broadcast to Enduro/X event server which then delivers the notification of record removal to all machines. The event is not sent in case if duplicate record is removed, as duplicate scan can be done by tpcached on other machine.
KEYGROUP OPERATIONS
tpcached can process any database. This can be normal service data storage databases or keygroup storage/keyitems storage - used for linking some kind of records into single group (for example user is doing paging over its statement records). Thus for example the group is user id and items are pages. In this case expiry and other limits can be applied to keygroup or key items databases too. In case if key item (some page expires) for db with keyitems flag, then the item is removed from group too. In case if group expires (database with keygroup flag), then all linked records are removed from keyitems database and the group record by it self is removed too. In case of group expire, only group key is broadcast to other nodes (if configured so).