Track unconverted device list outbound pokes using a position instead

[squahtx]

When a local device list change is added to
device_lists_changes_in_room, the converted_to_destinations flag is
set to FALSE and the _handle_new_device_update_async background
process is started. This background process looks for unconverted rows
in device_lists_changes_in_room, copies them to
device_lists_outbound_pokes and updates the flag.

To update the converted_to_destinations flag, the database performs a
DELETE and INSERT internally, which fragments the table. To avoid
this, track unconverted rows using a (stream ID, room ID) position
instead of the flag.

From now on, the converted_to_destinations column indicates rows that
need converting to outbound pokes, but does not indicate whether the
conversion has already taken place.

Closes #14037.

Signed-off-by: Sean Quah seanq@matrix.org

[squahtx]

We now do queries ordering by stream_id, room_id.
Shall we try to update the index to include room_id?

[erikjohnston]

Why are we removing these? Is these baths no longer used?

[squahtx]

Yes, pretty much. After removing the update to converted_to_destinations, we have

    async def add_device_list_outbound_pokes(
        self,
        user_id: str,
        device_id: str,
        room_id: str,
        hosts: Collection[str],
        context: Optional[Dict[str, str]],
    ) -> None:
        """Queue the device update to be sent to the given set of hosts,
        calculated from the room ID.
        """

        def add_device_list_outbound_pokes_txn(
            txn: LoggingTransaction, stream_ids: List[int]
        ) -> None:
            if hosts:
                self._add_device_outbound_poke_to_stream_txn(
                    txn,
                    user_id=user_id,
                    device_id=device_id,
                    hosts=hosts,
                    stream_ids=stream_ids,
                    context=context,
                )

        if not hosts:
            # If there are no hosts then we don't try and generate stream IDs.
            return await self.db_pool.runInteraction(
                "add_device_list_outbound_pokes",
                add_device_list_outbound_pokes_txn,
                [],
            )

        async with self._device_list_id_gen.get_next_mult(len(hosts)) as stream_ids:
            return await self.db_pool.runInteraction(
                "add_device_list_outbound_pokes",
                add_device_list_outbound_pokes_txn,
                stream_ids,
            )

The if not hosts branch does nothing inside the transaction, so I removed it entirely.

[erikjohnston]

I checked the queries against matrix.org DB and they seem sane