[ovs-dev] [PATCH] ovsdb: Implement garbage collection.

[Ben Pfaff]

Done, pushed.

On Thu, Mar 10, 2011 at 11:13:30AM -0800, Ethan Jackson wrote:
> Oh of course please run make check before merging.
> 
> Ethan
> 
> On Thu, Mar 10, 2011 at 11:11 AM, Ethan Jackson <ethan at nicira.com> wrote:
> > Thanks for writing these up.  Both incrementals look good.  Two very
> > minor aesthetic comments.  Go ahead and push.
> >
> > +    If "isRoot" is omitted or specified as false, then any given row
> > +    in the table may exist only when it there is at least one
> > "only when there is"
> >
> > -    def to_json(self):
> > +    def to_json(self, default_is_root=False):
> > +        """Returns this table schema serialized into JSON.
> > +
> > +        The "isRoot" member is included in the JSON only if its value would
> > Trailing whitespace here.
> >
> > Ethan
> >
> > On Wed, Mar 9, 2011 at 3:56 PM, Ben Pfaff <blp at nicira.com> wrote:
> >> Here's an additional incremental that should be applied on top.  I
> >> haven't tested it yet.
> >>
> >> diff --git a/debian/openvswitch-switch.init b/debian/openvswitch-switch.init
> >> index 92ab775..8ea5866 100755
> >> --- a/debian/openvswitch-switch.init
> >> +++ b/debian/openvswitch-switch.init
> >> @@ -232,6 +232,18 @@ case "$1" in
> >>             cksum=`ovsdb-tool db-cksum "$conf_file" | awk '{print $1}'`
> >>             cp "$conf_file" "$conf_file.backup$version-$cksum"
> >>
> >> +            # Compact database.  This is important if the old schema did not
> >> +            # enable garbage collection (i.e. if it did not have any tables
> >> +            # with "isRoot": true) but the new schema does.  In that situation
> >> +            # the old database may contain a transaction that creates a record
> >> +            # followed by a transaction that creates the first use of the
> >> +            # record.  Replaying that series of transactions against the new
> >> +            # database schema (as "convert" does) would cause the record to be
> >> +            # dropped by the first transaction, then the second transaction
> >> +            # would cause a referential integrity failure (for a strong
> >> +            # reference).
> >> +            ovsdb-tool -vANY:console:emer compact $conf_file
> >> +
> >>             # Upgrade or downgrade schema and compact database.
> >>             ovsdb-tool -vANY:console:emer convert $conf_file $schema_file
> >>         fi
> >> diff --git a/xenserver/etc_init.d_openvswitch b/xenserver/etc_init.d_openvswitch
> >> index 13b9d40..7300981 100755
> >> --- a/xenserver/etc_init.d_openvswitch
> >> +++ b/xenserver/etc_init.d_openvswitch
> >> @@ -341,7 +341,18 @@ function start {
> >>         cksum=`$ovsdb_tool db-cksum "$OVSDB_SERVER_DB" | awk '{print $1}'`
> >>         cp "$OVSDB_SERVER_DB" "$OVSDB_SERVER_DB.backup$version-$cksum"
> >>
> >> -        # Upgrade or downgrade schema and compact database.
> >> +        # Compact database.  This is important if the old schema did not enable
> >> +        # garbage collection (i.e. if it did not have any tables with "isRoot":
> >> +        # true) but the new schema does.  In that situation the old database
> >> +        # may contain a transaction that creates a record followed by a
> >> +        # transaction that creates the first use of the record.  Replaying that
> >> +        # series of transactions against the new database schema (as "convert"
> >> +        # does) would cause the record to be dropped by the first transaction,
> >> +        # then the second transaction would cause a referential integrity
> >> +        # failure (for a strong reference).
> >> +        $ovsdb_tool -vANY:console:emer compact "$OVSDB_SERVER_DB"
> >> +
> >> +        # Upgrade or downgrade schema.
> >>         $ovsdb_tool -vANY:console:emer convert "$OVSDB_SERVER_DB" "$VSWITCHD_OVSDB_SCHEMA"
> >>     fi
> >>
> >>
> >> _______________________________________________
> >> dev mailing list
> >> dev at openvswitch.org
> >> http://openvswitch.org/mailman/listinfo/dev
> >>
> >