Matrix-Mirrors/anonymousland-synapse
1
0
Fork 0
mirror of https://git.anonymousland.org/anonymousland/synapse.git synced 2024-10-01 11:49:51 -04:00
Code Issues Packages Projects Releases Wiki Activity

when we're talking about backfilling data in federation, call it backfilling - not pagination.

Browse Source
This commit is contained in:
Matthew Hodgson 2014-08-13 18:14:37 +01:00
parent f729f13735
commit 59dfbaba3b
9 changed files with 54 additions and 54 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
docs/python_architecture.rst
View File

@ -24,7 +24,7 @@ Where the bottom (the transport layer) is what talks to the internet via HTTP, a
* duplicate pdu_id's - i.e., it makes sure we ignore them.
* responding to requests for a given pdu_id
* responding to requests for all metadata for a given context (i.e. room)
* handling incoming pagination requests
* handling incoming backfill requests
So it has to parse incoming messages to discover which are metadata and which aren't, and has to correctly clobber existing metadata where appropriate.

4
docs/server-server/specification.rst
View File

@ -155,9 +155,9 @@ To fetch all the state of a given context:
PDUs that encode the state.
To paginate events on a given context:
To backfill events on a given context:
GET .../paginate/:context/
GET .../backfill/:context/
Query args: v, limit
Response: JSON encoding of a single Transaction containing multiple PDUs

6
docs/versioning.rst
View File

@ -1,11 +1,11 @@
Versioning is, like, hard for paginating backwards because of the number of Home Servers involved.
Versioning is, like, hard for backfilling backwards because of the number of Home Servers involved.
The way we solve this is by doing versioning as an acyclic directed graph of PDUs. For pagination purposes, this is done on a per context basis.
The way we solve this is by doing versioning as an acyclic directed graph of PDUs. For backfilling purposes, this is done on a per context basis.
When we send a PDU we include all PDUs that have been received for that context that hasn't been subsequently listed in a later PDU. The trivial case is a simple list of PDUs, e.g. A <- B <- C. However, if two servers send out a PDU at the same to, both B and C would point at A - a later PDU would then list both B and C.
Problems with opaque version strings:
- How do you do clustering without mandating that a cluster can only have one transaction in flight to a given remote home server at a time.
If you have multiple transactions sent at once, then you might drop one transaction, receive another with a version that is later than the dropped transaction and which point ARGH WE LOST A TRANSACTION.
- How do you do pagination? A version string defines a point in a stream w.r.t. a single home server, not a point in the context.
- How do you do backfilling? A version string defines a point in a stream w.r.t. a single home server, not a point in the context.
We only need to store the ends of the directed graph, we DO NOT need to do the whole one table of nodes and one of edges.

12
experiments/test_messaging.py
View File

@ -104,12 +104,12 @@ class InputOutput(object):
#self.print_line("OK.")
return
m = re.match("^paginate (\S+)$", line)
m = re.match("^backfill (\S+)$", line)
if m:
# we want to paginate a room
# we want to backfill a room
room_name, = m.groups()
self.print_line("paginate %s" % room_name)
self.server.paginate(room_name)
self.print_line("backfill %s" % room_name)
self.server.backfill(room_name)
return
self.print_line("Unrecognized command")
@ -307,7 +307,7 @@ class HomeServer(ReplicationHandler):
except Exception as e:
logger.exception(e)
def paginate(self, room_name, limit=5):
def backfill(self, room_name, limit=5):
room = self.joined_rooms.get(room_name)
if not room:
@ -315,7 +315,7 @@ class HomeServer(ReplicationHandler):
dest = room.oldest_server
return self.replication_layer.paginate(dest, room_name, limit)
return self.replication_layer.backfill(dest, room_name, limit)
def _get_room_remote_servers(self, room_name):
return [i for i in self.joined_rooms.setdefault(room_name,).servers]

4
synapse/federation/handler.py
View File

@ -75,8 +75,8 @@ class FederationEventHandler(object):
@log_function
@defer.inlineCallbacks
def backfill(self, room_id, limit):
# TODO: Work out which destinations to ask for pagination
# self.replication_layer.paginate(dest, room_id, limit)
# TODO: Work out which destinations to ask for backfill
# self.replication_layer.backfill(dest, room_id, limit)
pass
@log_function

6
synapse/federation/persistence.py
View File

@ -114,14 +114,14 @@ class PduActions(object):
@defer.inlineCallbacks
@log_function
def paginate(self, context, pdu_list, limit):
def backfill(self, context, pdu_list, limit):
""" For a given list of PDU id and origins return the proceeding
`limit` `Pdu`s in the given `context`.
Returns:
Deferred: Results in a list of `Pdu`s.
"""
results = yield self.store.get_pagination(
results = yield self.store.get_backfill(
context, pdu_list, limit
)
@ -131,7 +131,7 @@ class PduActions(object):
def is_new(self, pdu):
""" When we receive a `Pdu` from a remote home server, we want to
figure out whether it is `new`, i.e. it is not some historic PDU that
we haven't seen simply because we haven't paginated back that far.
we haven't seen simply because we haven't backfilled back that far.
Returns:
Deferred: Results in a `bool`

18
synapse/federation/replication.py
View File

@ -118,7 +118,7 @@ class ReplicationLayer(object):
*Note:* The home server should always call `send_pdu` even if it knows
that it does not need to be replicated to other home servers. This is
in case e.g. someone else joins via a remote home server and then
paginates.
backfills.
TODO: Figure out when we should actually resolve the deferred.
@ -179,13 +179,13 @@ class ReplicationLayer(object):
@defer.inlineCallbacks
@log_function
def paginate(self, dest, context, limit):
def backfill(self, dest, context, limit):
"""Requests some more historic PDUs for the given context from the
given destination server.
Args:
dest (str): The remote home server to ask.
context (str): The context to paginate back on.
context (str): The context to backfill.
limit (int): The maximum number of PDUs to return.
Returns:
@ -193,16 +193,16 @@ class ReplicationLayer(object):
"""
extremities = yield self.store.get_oldest_pdus_in_context(context)
logger.debug("paginate extrem=%s", extremities)
logger.debug("backfill extrem=%s", extremities)
# If there are no extremeties then we've (probably) reached the start.
if not extremities:
return
transaction_data = yield self.transport_layer.paginate(
transaction_data = yield self.transport_layer.backfill(
dest, context, extremities, limit)
logger.debug("paginate transaction_data=%s", repr(transaction_data))
logger.debug("backfill transaction_data=%s", repr(transaction_data))
transaction = Transaction(**transaction_data)
@ -281,9 +281,9 @@ class ReplicationLayer(object):
@defer.inlineCallbacks
@log_function
def on_paginate_request(self, context, versions, limit):
def on_backfill_request(self, context, versions, limit):
pdus = yield self.pdu_actions.paginate(context, versions, limit)
pdus = yield self.pdu_actions.backfill(context, versions, limit)
defer.returnValue((200, self._transaction_from_pdus(pdus).get_dict()))
@ -427,7 +427,7 @@ class ReplicationLayer(object):
# Get missing pdus if necessary.
is_new = yield self.pdu_actions.is_new(pdu)
if is_new and not pdu.outlier:
# We only paginate backwards to the min depth.
# We only backfill backwards to the min depth.
min_depth = yield self.store.get_min_depth_for_context(pdu.context)
if min_depth and pdu.depth > min_depth:

32
synapse/federation/transport.py
View File

@ -112,7 +112,7 @@ class TransportLayer(object):
return self._do_request_for_transaction(destination, subpath)
@log_function
def paginate(self, dest, context, pdu_tuples, limit):
def backfill(self, dest, context, pdu_tuples, limit):
""" Requests `limit` previous PDUs in a given context before list of
PDUs.
@ -126,14 +126,14 @@ class TransportLayer(object):
Deferred: Results in a dict received from the remote homeserver.
"""
logger.debug(
"paginate dest=%s, context=%s, pdu_tuples=%s, limit=%s",
"backfill dest=%s, context=%s, pdu_tuples=%s, limit=%s",
dest, context, repr(pdu_tuples), str(limit)
)
if not pdu_tuples:
return
subpath = "/paginate/%s/" % context
subpath = "/backfill/%s/" % context
args = {"v": ["%s,%s" % (i, o) for i, o in pdu_tuples]}
args["limit"] = limit
@ -251,8 +251,8 @@ class TransportLayer(object):
self.server.register_path(
"GET",
re.compile("^" + PREFIX + "/paginate/([^/]*)/$"),
lambda request, context: self._on_paginate_request(
re.compile("^" + PREFIX + "/backfill/([^/]*)/$"),
lambda request, context: self._on_backfill_request(
context, request.args["v"],
request.args["limit"]
)
@ -352,7 +352,7 @@ class TransportLayer(object):
defer.returnValue(data)
@log_function
def _on_paginate_request(self, context, v_list, limits):
def _on_backfill_request(self, context, v_list, limits):
if not limits:
return defer.succeed(
(400, {"error": "Did not include limit param"})
@ -362,7 +362,7 @@ class TransportLayer(object):
versions = [v.split(",", 1) for v in v_list]
return self.request_handler.on_paginate_request(
return self.request_handler.on_backfill_request(
context, versions, limit)
@ -371,14 +371,14 @@ class TransportReceivedHandler(object):
"""
def on_incoming_transaction(self, transaction):
""" Called on PUT /send/<transaction_id>, or on response to a request
that we sent (e.g. a pagination request)
that we sent (e.g. a backfill request)
Args:
transaction (synapse.transaction.Transaction): The transaction that
was sent to us.
Returns:
twisted.internet.defer.Deferred: A deferred that get's fired when
twisted.internet.defer.Deferred: A deferred that gets fired when
the transaction has finished being processed.
The result should be a tuple in the form of
@ -438,14 +438,14 @@ class TransportRequestHandler(object):
def on_context_state_request(self, context):
""" Called on GET /state/<context>/
Get's hit when someone wants all the *current* state for a given
Gets hit when someone wants all the *current* state for a given
contexts.
Args:
context (str): The name of the context that we're interested in.
Returns:
twisted.internet.defer.Deferred: A deferred that get's fired when
twisted.internet.defer.Deferred: A deferred that gets fired when
the transaction has finished being processed.
The result should be a tuple in the form of
@ -457,15 +457,15 @@ class TransportRequestHandler(object):
"""
pass
def on_paginate_request(self, context, versions, limit):
""" Called on GET /paginate/<context>/?v=...&limit=...
def on_backfill_request(self, context, versions, limit):
""" Called on GET /backfill/<context>/?v=...&limit=...
Get's hit when we want to paginate backwards on a given context from
Gets hit when we want to backfill backwards on a given context from
the given point.
Args:
context (str): The context to paginate on
versions (list): A list of 2-tuple's representing where to paginate
context (str): The context to backfill
versions (list): A list of 2-tuples representing where to backfill
from, in the form `(pdu_id, origin)`
limit (int): How many pdus to return.

20
synapse/storage/pdu.py
View File

@ -168,7 +168,7 @@ class PduStore(SQLBaseStore):
return self._get_pdu_tuples(txn, txn.fetchall())
def get_pagination(self, context, pdu_list, limit):
def get_backfill(self, context, pdu_list, limit):
"""Get a list of Pdus for a given topic that occured before (and
including) the pdus in pdu_list. Return a list of max size `limit`.
@ -182,12 +182,12 @@ class PduStore(SQLBaseStore):
list: A list of PduTuples
"""
return self._db_pool.runInteraction(
self._get_paginate, context, pdu_list, limit
self._get_backfill, context, pdu_list, limit
)
def _get_paginate(self, txn, context, pdu_list, limit):
def _get_backfill(self, txn, context, pdu_list, limit):
logger.debug(
"paginate: %s, %s, %s",
"backfill: %s, %s, %s",
context, repr(pdu_list), limit
)
@ -213,7 +213,7 @@ class PduStore(SQLBaseStore):
new_front = []
for pdu_id, origin in front:
logger.debug(
"_paginate_interaction: i=%s, o=%s",
"_backfill_interaction: i=%s, o=%s",
pdu_id, origin
)
@ -224,7 +224,7 @@ class PduStore(SQLBaseStore):
for row in txn.fetchall():
logger.debug(
"_paginate_interaction: got i=%s, o=%s",
"_backfill_interaction: got i=%s, o=%s",
*row
)
new_front.append(row)
@ -262,7 +262,7 @@ class PduStore(SQLBaseStore):
def update_min_depth_for_context(self, context, depth):
"""Update the minimum `depth` of the given context, which is the line
where we stop paginating backwards on.
on which we stop backfilling backwards.
Args:
context (str)
@ -320,9 +320,9 @@ class PduStore(SQLBaseStore):
return [(row[0], row[1], row[2]) for row in results]
def get_oldest_pdus_in_context(self, context):
"""Get a list of Pdus that we paginated beyond yet (and haven't seen).
This list is used when we want to paginate backwards and is the list we
send to the remote server.
"""Get a list of Pdus that we haven't backfilled beyond yet (and haven't
seen). This list is used when we want to backfill backwards and is the
list we send to the remote server.
Args:
txn