launchpad-reviewers team mailing list archive

[Stuart Bishop <stuart.bishop@xxxxxxxxxxxxx>]

Stuart Bishop has proposed merging lp:~stub/launchpad/staging into lp:launchpad.

Requested reviews:
  Launchpad code reviewers (launchpad-reviewers)

For more details, see:
https://code.launchpad.net/~stub/launchpad/staging/+merge/110818

= Summary =

We are switching from Slony to PostgreSQL 9.1 streaming replication and the staging rebuild scripts need to be refactored.

== Proposed fix ==

== Pre-implementation notes ==

== LOC Rationale ==

== Implementation details ==

== Tests ==

== Demo and Q/A ==


= Launchpad lint =

Checking for conflicts and issues in changed files.

Linting changed files:
  database/schema/upgrade.py
  database/replication/Makefile

./database/schema/upgrade.py
      12: '_pythonpath' imported but unused
-- 
https://code.launchpad.net/~stub/launchpad/staging/+merge/110818
Your team Launchpad code reviewers is requested to review the proposed merge of lp:~stub/launchpad/staging into lp:launchpad.

=== modified file 'database/replication/Makefile'
--- database/replication/Makefile	2012-04-24 14:11:52 +0000
+++ database/replication/Makefile	2012-06-18 13:30:16 +0000
@@ -1,21 +1,15 @@
 # Copyright 2009-2012 Canonical Ltd.  This software is licensed under the
 # GNU Affero General Public License version 3 (see the file LICENSE).
 #
-# This makefile is used for two task:
-#
-#  - Build a replicated development environment
-#  - Build and replicate the staging database
-#
-# To build a replicated development environment:
-#
-#  $ make -C database/replication devsetup
-#  $ make run
+# This makefile is used to build and replicate the staging database.
+# To build a replicated development environent, replicate your 9.1 main
+# cluster with standard PG 9.1 streaming replication per the PG docs.
 #
 # To test the staging rebuild script:
 #
 #  $ mkdir -p /srv/staging.launchpad.net/staging-logs
 #  $ cd database/replication
-#  $ pg_dump --exclude-schema=_sl --format=c launchpad_dev > launchpad.dump
+#  $ pg_dump --format=c launchpad_dev > launchpad.dump
 #  $ make stagingsetup \
 #        STAGING_CONFIG=dev-staging STAGING_DUMP=launchpad.dump
 #  $ make stagingswitch STAGING_CONFIG=dev-staging
@@ -26,18 +20,10 @@
 #  $ make dogfood DOGFOOD_DBNAME=launchpad_dogfood DOGFOOD_DUMP=launchpad.dump
 #
 
-# This used to be 10 seconds, so we always ran staging lagged to detect
-# replication glitches more easily. However, this does not play well
-# with DBLoopTuner, as it correctly blocks when it detects lag.
-# We should put this back when DBLoopTuner is more intelligent, perhaps
-# pulling the allowable lag setting from a config file.
-LAG=0 seconds
-
 # For real restores, this is being run on an 8 core system. Let pg_restore
 # run multiprocess.
 MULTIPROC=-j 4
 
-DEV_CONFIG=replicated-development
 NEW_STAGING_CONFIG=staging-setup # For building the db with a different name.
 STAGING_CONFIG=staging-db # For swapping fresh db into place.
 STAGING_DUMP=launchpad.dump # Dumpfile to build new staging from.
@@ -46,18 +32,9 @@
 DOGFOOD_DBNAME=launchpad_dogfood
 DOGFOOD_DUMP=launchpad.dump
 
-_CONFIG=overridden-on-command-line
-_SLAVE_TABLESPACE=pg_default
-
 PGMASSACRE=../../utilities/pgmassacre.py
 
-CREATEDB_83=createdb --encoding=UTF8
-CREATEDB_84=createdb --encoding=UTF8 --locale=C --template=template0
-CREATEDB=${CREATEDB_84}
-
-# Set this to --exit-on-error once we have migrated Launchpad production
-# to Slony-I 2.0.x.
-EXIT_ON_ERROR=
+CREATEDB=createdb --encoding=UTF8 --locale=C --template=template0
 
 # Turn off output silencing so we can see details of staging deployments.
 # Without the timestamps, we are unable to estimate production deployment
@@ -66,161 +43,39 @@
 SHHH=
 
 default:
-	echo Usage: make [start|stop|restart]
-
-start:
-	./slon_ctl.py --lag="${LAG}" start
-
-stop:
-	./slon_ctl.py stop
-
-restart: stop start
-
-devsetup:
-	make _prelim LPCONFIG=${DEV_CONFIG} \
-	    _MASTER=launchpad_dev _SLAVE=launchpad_dev_slave LAG="0 seconds"
-	
-	# Build the master database
-	LPCONFIG=${DEV_CONFIG} make -C ../schema
-	
-	# Replicate it
-	make _replicate LPCONFIG=${DEV_CONFIG} \
-	    _MASTER=launchpad_dev _SLAVE=launchpad_dev_slave LAG="0 seconds"
-	
-	# Replicate it again, so we can test with multiple slaves.
-	-${PGMASSACRE} launchpad_dev_slave2
-	${CREATEDB} launchpad_dev_slave2
-	LPCONFIG=${DEV_CONFIG} ./slon_ctl.py start \
-		 node3_node 'dbname=launchpad_dev_slave2 user=slony'
-	LPCONFIG=${DEV_CONFIG} ./new-slave.py 3 launchpad_dev_slave2
-	LPCONFIG=${DEV_CONFIG} ../schema/security.py -d launchpad_dev_slave2
-	
-	# Regenerate the preamble for manual slonik(1) usage.
-	LPCONFIG=${DEV_CONFIG} ./preamble.py > preamble.sk
-
-	# Restart slon daemons with default lag setting.
-	LPCONFIG=${DEV_CONFIG} ./slon_ctl.py stop
-	LPCONFIG=${DEV_CONFIG} ./slon_ctl.py --lag="${LAG}" start
-
+	echo Need a target
 
 # Build _new staging databases from a production dump.
-stagingsetup: DUMPLIST:=$(shell tempfile --suffix=.lst --prefix=lpstag)
 stagingsetup:
-	make _prelim LPCONFIG=${NEW_STAGING_CONFIG} \
-	    _MASTER=lpmain_staging_new _SLAVE=lpmain_staging_slave_new \
-	    LAG="0 seconds"
+	# Kill the existing temporary staging database if it exists from an
+	# aborted run
+	-${PGMASSACRE} lpmain_staging_new
 	# Create the DB with the desired default tablespace.
 	${CREATEDB} --tablespace ${STAGING_TABLESPACE} lpmain_staging_new
-	# Drop the plpgsql language if it was installed, as it will be
-	# recreated with the dump from PostgreSQL 8.4
-	-droplang plpgsql lpmain_staging_new
 	# Restore the database. We need to restore permissions, despite
 	# later running security.py, to pull in permissions granted on
 	# production to users not maintained by security.py.
-	# Unfortunately, we need to dump production databases using
-	# --exclude-schema=_sl, or the read locks opened by pg_dump will
-	# cause the replication to appear lagged and set off alerts.
-	# This makes the restoration process more complex - a human being
-	# can just restore the dump, scan the errors and confirm it worked.
-	# For this automation, we need to massage the list of data to
-	# restore from the dump to avoid the spurious errors and get a
-	# valid error code.
-	pg_restore --list ${STAGING_DUMP} | grep -v 'TRIGGER public _sl_' \
-	    > ${DUMPLIST}
-	pg_restore --dbname=lpmain_staging_new --no-owner ${EXIT_ON_ERROR} \
-	    --use-list=${DUMPLIST} ${MULTIPROC} ${STAGING_DUMP}
-	rm ${DUMPLIST}
-	# Uninstall Slony-I if it is installed - a pg_dump of a DB with
-	# Slony-I installed isn't usable without this step. This is likely
-	# irrelevant, as if our dump was created with --exclude-schema=_sl
-	# it will not contain anything to remove.
-	LPCONFIG=${NEW_STAGING_CONFIG} ./repair-restored-db.py
-	# Apply database patches now. This is unfortunate, as it means
-	# the patches are not being applied to a replicated database.
-	# With this update, we get incorrect patch application times and
-	# won't catch patches incompatible with Slony. We do this now,
-	# because we are running short of disk space on the staging
-	# database server. With the patches being applied before
-	# replication, we can swap the slave database into place before
-	# it has finished building and we end up using only 3*production
-	# amount of disk space rather than 4*production.
-	@echo Running upgrade.py `date` non-replicated. Times need adjustment.
+	pg_restore --dbname=lpmain_staging_new ${EXIT_ON_ERROR} \
+	    ${MULTIPROC} ${STAGING_DUMP}
+	# Apply database patches.
+	@echo Running upgrade.py `date`.
 	LPCONFIG=${NEW_STAGING_CONFIG} ${SHHH} ../schema/upgrade.py \
-	    --log-file=INFO:${STAGING_LOGDIR}/dbupgrade.log
+	    --ignore-slony --log-file=INFO:${STAGING_LOGDIR}/dbupgrade.log
 	@echo Running security.py `date`
-	LPCONFIG=${NEW_STAGING_CONFIG} ${SHHH} ../schema/security.py -U slony \
+	LPCONFIG=${NEW_STAGING_CONFIG} ${SHHH} ../schema/security.py \
 	    --log-file=INFO:${STAGING_LOGDIR}/dbupgrade.log
-	# Setup replication
-	make _replicate LPCONFIG=${NEW_STAGING_CONFIG} LAG="0 seconds" \
-	    _MASTER=lpmain_staging_new _SLAVE=lpmain_staging_slave_new \
-	    _SLAVE_TABLESPACE=${STAGING_TABLESPACE}
 
 # Switch the _new staging databases into place.
 stagingswitch:
-	# Stop Slony-I daemons - don't confuse the poor darlings.
-	-LPCONFIG=${NEW_STAGING_CONFIG} ./slon_ctl.py stop
-	-LPCONFIG=${STAGING_CONFIG} ./slon_ctl.py stop
 	# Kill the existing staging database if it exists.
 	-${PGMASSACRE} lpmain_staging
-	-${PGMASSACRE} lpmain_staging_slave
 	# Rename the newly build staging databases.
 	psql -d template1 -c \
 	    "ALTER DATABASE lpmain_staging_new RENAME TO lpmain_staging;"
-	psql -d template1 -c "\
-	    ALTER DATABASE lpmain_staging_slave_new \
-	    RENAME TO lpmain_staging_slave;"
-	# Fix the paths to match.
-	psql -d lpmain_staging -U slony -c \
-	"UPDATE _sl.sl_path SET pa_conninfo=replace(pa_conninfo, '_new', '')"
-	psql -d lpmain_staging_slave -U slony -c \
-	"UPDATE _sl.sl_path SET pa_conninfo=replace(pa_conninfo, '_new', '')"
-	# Start the slon daemons, with requested lag.
-	LPCONFIG=${STAGING_CONFIG} ./slon_ctl.py --lag="${LAG}" start
 
-dogfood: DUMPLIST:=$(shell tempfile --suffix=.lst --prefix=lpdogf)
 dogfood:
 	${CREATEDB} ${DOGFOOD_DBNAME}
-	# May be needed with PG 9.1 if the dump is from PG 8.4.
-	-droplang plpgsql ${DOGFOOD_DBNAME}
-	pg_restore --list ${DOGFOOD_DUMP} | grep -v 'TRIGGER public _sl_' \
-	    > ${DUMPLIST}
 	pg_restore --dbname=${DOGFOOD_DBNAME} --no-acl --no-owner \
-	    --use-list=${DUMPLIST} ${EXIT_ON_ERROR} ${DOGFOOD_DUMP}
-	rm ${DUMPLIST}
-	./repair-restored-db.py -d ${DOGFOOD_DBNAME}
-	../schema/upgrade.py -d ${DOGFOOD_DBNAME}
+	    ${EXIT_ON_ERROR} ${DOGFOOD_DUMP}
+	../schema/upgrade.py --ignore-slony -d ${DOGFOOD_DBNAME}
 	../schema/security.py -d ${DOGFOOD_DBNAME}
-
-_prelim:
-	@echo LPCONFIG currently ${LPCONFIG}
-	# Create the slony PostgreSQL superuser if necessary.
-	-createuser --superuser slony
-	# Stop the slon daemons and wait a bit for connections to drop.
-	-./slon_ctl.py stop
-	sleep 5
-	# Drop any existing databases if they exist
-	${PGMASSACRE} ${_MASTER}
-	${PGMASSACRE} ${_SLAVE}
-
-_replicate:
-	@echo LPCONFIG currently ${LPCONFIG}
-	# Start the slon daemon for the master.
-	./slon_ctl.py --lag="0 seconds" start \
-		 node1_node "dbname=${_MASTER} user=slony"
-	# Initialize the cluster and create replication sets.
-	./initialize.py \
-	    --log-file=INFO:${STAGING_LOGDIR}/dbupgrade.log
-	# Create the soon-to-be-slave database, empty at this point.
-	${CREATEDB} --tablespace=${_SLAVE_TABLESPACE} ${_SLAVE}
-	# Start the slon daemon for the slave
-	./slon_ctl.py --lag="0 seconds" start \
-	    node2_node "dbname=${_SLAVE} user=slony"
-	# Setup the slave
-	./new-slave.py 2 "dbname=${_SLAVE}" \
-	    --log-file=INFO:${STAGING_LOGDIR}/dbupgrade.log
-	# Restart slon daemons with default lag setting.
-	./slon_ctl.py stop
-	./slon_ctl.py --lag="${LAG}" start
-	# Generate a preamble for manual slonik(1) usage.
-	./preamble.py > preamble.sk
-

=== modified file 'database/schema/upgrade.py'
--- database/schema/upgrade.py	2012-05-31 09:54:57 +0000
+++ database/schema/upgrade.py	2012-06-18 13:30:16 +0000
@@ -44,7 +44,7 @@
     con = connect()
     patches = get_patchlist(con)
 
-    if replication.helpers.slony_installed(con):
+    if not options.ignore_slony and replication.helpers.slony_installed(con):
         con.close()
         if options.commit is False:
             parser.error("--dry-run does not make sense with replicated db")
@@ -609,14 +609,14 @@
     db_options(parser)
     logger_options(parser)
     parser.add_option(
-            "-n", "--dry-run", dest="commit", default=True,
-            action="store_false", help="Don't actually commit changes"
-            )
-    parser.add_option(
-            "--partial", dest="partial", default=False,
-            action="store_true",
-            help="Commit after applying each patch",
-            )
+        "-n", "--dry-run", dest="commit", default=True,
+        action="store_false", help="Don't actually commit changes")
+    parser.add_option(
+        "--partial", dest="partial", default=False,
+        action="store_true", help="Commit after applying each patch",)
+    parser.add_option(
+        "--ignore-slony", dest="ignore_slony", default=False,
+        action="store_true", help="Ignore any Slony installations",)
     (options, args) = parser.parse_args()
 
     if args: