--- /dev/null
+==== requirements
+you need to have
+(*) one master test machine (abbreved 'test') - this is where you
+trigger everything from
+(*) two servers with myplc installed (abbreved 'plc1' and 'plc2')
+at this stage the myplc configuration needs to be done by hand
+you also need root ssh access from the test box to the two plc boxes in both boxes
+
+-- configuring the plc servers:
+(*) set their names in the Makefile (PLC1 and PLC2)
+(*) edit TestPeers.py too -- xxx should be improved
+
+To some extent this stuff can be ran to control a single plc.
+
+==== concepts
+right now, 4 tests sizes are supported, they are named
+* m (minimal) 1 object of each kind is created
+* n (normal) a few of each kind
+* b (big) a few hundreds
+* h (huge) a few thousands
+
+two modes are supported:
+
+(1) single run mode: everything is ran from the 'test' box
+this is a convenient, but very slow mode, especially for large sizes
+because each individual object is created in a single xmlrpc command
+>>> xxx btw, I did not know about begin/commit at that time, but I do not
+ think it can speed up to the point where the second mode does
+In this mode, both DB's are populated together, and various checks can
+be done on the fly
+
+(2) populate-and-run mode
+where the DB populating parts are done beforehand, and separately,
+from each plc's chroot jail for direct access to the DB
+This mode allows for dump&restore of the populated DB
+
+==== preparation
+
+(*) manually install both myplc, do the configs manually (once and for
+good, see upgrades below)
+
+(*) initial setup
+- check Makefile and TestPeers.py for correct identification of
+
+(*) push the stuff
+$ make push
+so the local repository gets synced on all nodes.
+
+(*) get and push peers information
+$ make peers
+(at least once) so the gpg and other authentication materials get pushed to both nodes
+
+==== use cases
+
+IMPORTANT: for cleaning up previous tests if needed
+
+$ make db-clean.3
+=> cleans up both databases if needed
+<<< xxx this uses the initscripts to perform the job, it's a bit slow on
+old boxes, could probably be improved by snapshotting the db right
+after plc gets started>>>
+
+(1)
+
+$ make testpeers.help
+
+$ make testpeers-m.run
+assumes the dbs are clean, and runs the test locally
+
+$ make testpeers-b.all
+cleans both dbs and runs .run
+
+$ make testpeers-m.diff
+checks the output agaist the .ref file that chould be under subversion
+
+$ make testpeers-m.ckp
+adopts the current .out as a reference for that test - does *not*
+commit under subversion, just copies the .out into .ref
+
+(2)
+
+$ make populate.help
+
+$ make populate-m.all
+cleans both dbs, cleans any former result, then runs .init and .run
+
+$ make populate-b.init
+assumes both dbs are clean, populates both databases and dumps database in the .sql files
+
+$ make populate-b.run
+performs the populate_end part of the test from the populated database
+
+$ make populate-b.restore
+restores the database from the stage where it just populated
+
+$ make populate-b.clean
+cleans everything except the sql files
+
+$ make populate-b.sqlclean
+cleans the .sql files
+
+==== various utilities
+
+(*) upgrading both plcs (xxx could be improved):
+- log as root on both plc servers and manually curl the myplc rpm you
+want, in /root/new_plc_api/tests
+- back to the server node:
+$ make upgrade.3
+
+if you want to upgrade only one, of course just use targets upgrade.1
+or upgrade.2
+
+(*) cleaning the database
+$ make db-clean.1 (or .2 or .3)
+
+
+
+==== implementation notes
+I've designed this thing so everything can be invoked from the test
+server, even when things need to be actually triggered from a chroot
+
+For this reason the same pieces of code (namely Makefile and
+TestPeers.py) needs to be accessible from
+(*) the test server
+(*) root's homedir on both plc's, namely in /root/new_plc_api/tests
+(*) chroot jail on both plc's, namely in /plc/root/usr/share/plc_api/tests
+
+This is where the 'push' target comes in
+xxx at this stage the push target pushes the whole API, because I used this
+to test the code that I was patching on the test node. this can be improved
+
+TARGET structure
+
+If you need to invoke something on a given plc, append '.1' or '.2' to the target name
+$ make db-clean.1
+Will run the 'db-clean' target on the plc1 node
+
+for convenience:
+$ make db-clean.3
+will run db-clean on both plc's
+
+When something needs to be ran in the chroot jail, run
+$ make sometarget.chroot
+that will make sometarget from the chroot jail
+
+So from the test server
+$ make sometarget.chroot.3
+runs in both chroot jails
+
git://git.onelab.eu / plcapi.git / blobdiff