hvrswitchtable - Schedule switch of one channel's tables into another channel without interrupting replication.
hvrswitchtable [-options] hubdb chn1 chn2
hvrswitchtable [-m] [-i latency] [-k] [-k table]... [-h class] [-u user[/pwd]] hubdb chn1 chn2
hvrswitchtable [-a | -A integ_loc] [-s] [-T timeo] [-h class] [-u user[/pwd]] hubdb chn1 chn2
hvrswitchtable encapsulates a series of steps required to prepare a switch of some or all tables of channel chn1 into channel chn2 at a given 'switch moment'. This is a moment in the future which is relative to hvr_cap_tstamp value and not the job's processing time. Switching tables means moving all the specified tables including actions which are explicitly defined on chn1 and these tables (i.e. actions where channel is chn1 and table is not ‘*’) from chn1 into chn2. The prepared switch will be performed without interrupting replication of any table in chn1 and chn2.
hvrswitchtable can be used to add tables to an existing channel without having to stop capture or integrate.
Requirements for a successful channel merge are:
- All locations of chn1 and chn2 must be the same.
- The location groups of chn1 and chn2 must have the same names and members.
- Table names of the tables being switched from chn1 should not be used in chn2.
- All tables in chn1 must be refreshed.
- The capture and integrate jobs in chn1 and chn2 should be running with a low latency.
This section describes the options available for command hvrswitchtable.
Abort switch. Removes newly added tables and actions from chn2 and let chn1 continue with the replication. This is not possible if switch has already succeeded for one or more integrate locations. HVR Scheduler must be running to abort switch.
Abort switch only for specific integrate location.
Specify hub database. For supported values, see Calling HVR on the Command Line.
Specify an initial minimum latency. Switch if tables will not be performed if one of the involved jobs has a initial latency higher than latency. Valid inputs are an integer (in seconds) or a time duration (e.g. 10s, 3m, 1h).
Keep chn1 integrate jobs. Do not delete integrate jobs of chn1 after successful switch.
Specify future switch moment relative to hvr_cap_tstamp.
Do not suspend and restart integrate jobs when aborting.
Specify table scope. Only switch table table from chn1 to chn2. This option can be supplied multiple times. If not specified all tables will be switched and integrate jobs of chn1 will be deleted afterward (unless -k is specified). It is not possible to switch all tables using option -t.
Timeout for connection to HVR Scheduler. The default is 10 seconds.
Connect to hub database using DBMS account user. For some databases (e.g. SQL Server) a password must also be supplied.
Example 1: Adding tables to a running channel via a temporary channel
Consider a running channel chn_main replicating tables tab_main1 and tab_main2 from capture location cap to integrate location integ. Location cap is member of location group CAP and location integ is member of location group INTEG. chn_main is running with a low latency and without errors. Now hvrswitchtable can be used to add new tables tab_new1 and tab_new2 without interrupting capture or integrate. First a new channel chn_tmp is created with the same location groups CAP and INTEG. Then new tables tab_new1 and tab_new2 are added using hvradapt,chn_tmp is initialize using hvrinit and after starting the capture job the new tables are refreshed with an online hvrrefresh. After hvrrefresh completed successfully the integrate job can be started and HVR will start replicating all changes made to the new tables. After ensuring that chn_tmp runs fine with a low latency, a switch if the tables of chn_tmp into chn_main can be prepared.
Executing the following command will first check that all capture and integrate jobs in chn_tmp and chn_main run with a latency lower than 10 seconds. If so it adds tables tab_new1 and tab_new2 to chn_main and prepares to switch the tables in 1 minute.
Note that the 'now' in the command line refers to the current time (not relative to hvr_cap_tstamp). After chn_tmp has captured and integrated all changes that were made to tab_new1 and tab_new2 up to 1 minute after hvrswitchtable was run, chn_main will capture and integrate all future changes and the integrate job of chn_tmp will be deleted. Now all tables have been switched and chn_tmp can be deleted.
Example 2: Schedule and abort a merge of two channels
The scenario is similar to Example 1 but now location group INTEG contains a second integrate location intg2 and chn_tmp is already running fine with a low latency. In this example the current time is '2017-12-10T14:30:00+01:00'.
Executing the following command prepares the switch for '2017-12-10T15:00:00+01:00', i.e., in 30 minutes and the integrate jobs of chn_tmp will not be deleted after the switch:
If you want to abort the switch before '2017-12-10T15:00:00+01:00', execute the following command:
Aborting the switch removes tables tab_new1 and tab_new2 from chn_main and the replication of both channels will continue without any interruption.
Assume the switch was not aborted and at '2017-12-10T14:58:00+01:00' the integrate job for location intg2 in channel chn_tmp (chn_tmp-integ-intg2) failed due to some integration error. At '2017-12-10T15:00:00+01:00' tables have been switched successfully for location integ, but chn_tmp-integ-intg2 is still failing. The integrate job for intg2 in chn_main (chn_main-integ-intg2) will wait for chn_tmp-integ-intg2 to integrate all changes up to '2017-12-10T15:00:00+01:00', so now it is hanging. To abort the switch only for location intg2, execute the following command:
This will not remove tables tab_new1 and tab_new2 from chn_main, but will restart chn_main-integ-intg2 and ensure that chn_main will not replicate any changes for tab_new1 and tab_new2. After integration issues in chn_tmp have been resolved, integration of tab_new1 and tab_new2 into intg2 will continue in chn_tmp.
Example 3: Switch some tables between two running channels
Consider a situation where two identically configured channels chn_from and chn_to are moving tables at a low latency. For this example chn_from is replicating tables tab1, tab2 and tab3 and chn_to is replicating tables tab4 and tab5. To switch tables tab2 and tab3 from chn_from to chn_to in 1 hour execute the following command.
This will add tab2 and tab3 and all their associated actions to chn_to. After all involved integrate jobs are passed the given switch moment (i.e. hvr_cap_tstamp has gone passed now +1 hour), chn_to will replicate all changes for tab2 and tab3 and the tables and their associated actions will be removed from chn_from. That means chn_from will now only replicate tab1 and chn_to will replicate tab2, tab3, tab4 and tab5. No further actions are required and chn_from and chn_to will continue with their replication. Note that aborting of the switch works as in Example 2, only the table scope (option -t) must be provided for the abort command as well.