From Paweł:
This is another episode in the "convert X to streamed mutations" series.
Hashing mutations (mainly for repair) is converted so that it doesn't
need to rebuild whole mutation.
The first part of the series changes the way streamed mutations deal
with range tombstones. Since it is not necessary to make sure we write
disjoint tombstones to sstables there is no need anymore for streamed
mutations to produce disjoint tombstones and, consequently, no need for
range tombstones to be split into range_tombstone_begin and
range_tombstone_end.
The second part is the actual hashing implementation. However, to ensure
that the hash depends only on the contents of the mutation and no the
way it is stored in different data sources range tombstones have to be
made disjoint before they are hashed.
This series also ensures that any changes caused by streamed mutations
to hashing and streaming do not break repair during upgrade.
The receiving side needs to handle fragmented mutations properly so that
isolation guarantees are not broken. If the receiving node may be an old
one do not fragment mutations.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Commit 206955e4 "streaming: Reduce memory usage when sending mutations"
moved streaming mutation limiter from do_send_mutations() to
send_mutations(). The reason for that was that send_mutation() did full
mutation copies. That's no longer the case and streaming limiter should
be moved back to do_send_mutation() in order to provide back pressure to
fragment_and_freeze().
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
If mutations are fragmented during streaming a special care must be
taken so that isolation guarantees are not broken.
Mutations received with flag "fragmented" set are applied to a memtable
that is used only by that particular streaming task and the sstables
created by flushing such memtables are not made visible until the task
is complte. Also, in case the streaming fails all data is dropped.
This means that fragmented mutations cannot benefit from coalescing of
writes from multiple streaming plans, hence separate way of handling
them so that there is no loss of performance for small partitions.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
plan_id is needed to keep track of the origin of mutations so that if
they are fragmented all fragments are made visible at the same time,
when that particular streaming plan_id completes.
Basically, each streaming plan that sends big (fragmented) mutations is
going to have its own memtables and a list of sstables which will get
flushed and made visible when that plan completes (or dropped if it
fails).
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
dtest takes error level log as serious error. It is not a serious error
for streaming to fail to send a verb and fail a streaming session, for
example, the peer node is gone or stopped. Switch to use log level warn
instead of level error.
Fixes repair_additional_test.py:RepairAdditionalTest.repair_kill_3_test
Fixes: #1335
Message-Id: <0149d30044e6e4d80732f1a20cd20593de489fc8.1465979288.git.asias@scylladb.com>
Limit disk bandwidth to 5MB/s to emulate a slow disk:
echo "8:0 5000000" >
/cgroup/blkio/limit/blkio.throttle.write_bps_device
echo "8:0 5000000" >
/cgroup/blkio/limit/blkio.throttle.read_bps_device
Start scylla node 1 with low memory:
scylla -c 1 -m 128M --auto-bootstrap false
Run c-s:
taskset -c 7 cassandra-stress write duration=5m cl=ONE -schema
'replication(factor=1)' -pop seq=1..100000 -rate threads=20
limit=2000/s -node 127.0.0.1
Start scylla node 2 with low memory:
scylla -c 1 -m 128M --auto-bootstrap true
Without this patch, I saw std::bad_alloc during streaming
ERROR 2016-06-01 14:31:00,196 [shard 0] storage_proxy - exception during
mutation write to 127.0.0.1: std::bad_alloc (std::bad_alloc)
...
ERROR 2016-06-01 14:31:10,172 [shard 0] database - failed to move
memtable to cache: std::bad_alloc (std::bad_alloc)
...
To fix:
1. Apply the streaming mutation limiter before we read the mutation into
memory to avoid wasting memory holding the mutation which we can not
send.
2. Reduce the parallelism of sending streaming mutations. Before we send each
range in parallel, after we send each range one by one.
before: nr_vnode * nr_shard * (send_info + cf.make_reader memory usage)
after: nr_shard * (send_info + cf.make_reader memory usage)
We can at least save memory usage by the factor of nr_vnode, 256 by
default.
In my setup, fix 1) alone is not enough, with both fix 1) and 2), I saw
no std::bad_alloc. Also, I did not see streaming bandwidth dropped due
to 2).
In addition, I tested grow_cluster_test.py:GrowClusterTest.test_grow_3_to_4,
as described:
https://github.com/scylladb/scylla/issues/1270#issuecomment-222585375
With this patch, I saw no std::bad_alloc any more.
Fixes: #1270
Message-Id: <7703cf7a9db40e53a87f0f7b5acbb03fff2daf43.1464785542.git.asias@scylladb.com>
- Do nothing in case the session is closed, to prevent we fire up the
timer again
- Print log info when no progress has been made if the time expires, it
is very useful to debug a idle session
- Grab a reference when the keep alive timer is running
Message-Id: <9f2cc3164696905a6a39c0d072a980765d598dfd.1458782956.git.asias@scylladb.com>
A STREAM_MUTATION_DONE message will signal the receiver that the sender
has completed the sending of streams mutations. When the receiver finds
it has zero task to send and zero task to receive, it will finish the
stream_session, and in turn finish the stream_plan if all the
stream_sessions are finished. We should call receive_task_completed only
after the flush finishes so that when stream_plan is finshed all the
data is on disk.
Fixes repair_disjoint_data_test issue with Glauber's "[PATCH v4 0/9] Make
sure repairs do not cripple incoming load" serries
======================================================================
FAIL: repair_disjoint_data_test
(repair_additional_test.RepairAdditionalTest)
----------------------------------------------------------------------
Traceback (most recent call last):
File "scylla-dtest/repair_additional_test.py",
line 102, in repair_disjoint_data_test
self.check_rows_on_node(node1, 3000)
File "scylla-dtest/repair_additional_test.py",
line 33, in check_rows_on_node
self.assertEqual(len(result), rows, len(result))
AssertionError: 2461
Keeping the mutations coming from the streaming process as mutations like any
other have a number of advantages - and that's why we do it.
However, this makes it impossible for Seastar's I/O scheduler to differentiate
between incoming requests from clients, and those who are arriving from peers
in the streaming process.
As a result, if the streaming mutations consume a significant fraction of the
total mutations, and we happen to be using the disk at its limits, we are in no
position to provide any guarantees - defeating the whole purpose of the
scheduler.
To implement that, we'll keep a separate set of memtables that will contain
only streaming mutations. We don't have to do it this way, but doing so
makes life a lot easier. In particular, to write an SSTable, our API requires
(because the filter requires), that a good estimate on the number of partitions
is informed in advance. The partitions also need to be sorted.
We could write mutations directly to disk, but the above conditions couldn't be
met without significant effort. In particular, because mutations can be
arriving from multiple peer nodes, we can't really sort them without keeping a
staging area anyway.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Streaming has currently one class, that can be used to contain the read
operations being generated by the streaming process. Those reads come from two
places:
- checksums (if doing repair)
- reading mutations to be sent over the wire.
Depending on the amount of data we're dealing with, that can generate a
significant chunk of data, with seconds worth of backlog, and if we need to
have the incoming writes intertwined with those reads, those can take a long
time.
Even if one node is only acting as a receiver, it may still read a lot for the
checksums - if we're talking about repairs, those are coming from the
checksums.
However, in more complicated failure scenarios, it is not hard to imagine a
node that will be both sending and receiving a lot of data.
The best way to guarantee progress on both fronts, is to put both kinds of
operations into different classes.
This patch introduces a new write class, and rename the old read class so it
can have a more meaningful name.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Below are 3 possible cases in a stream session, after commit
208b7fa7ba (streaming: Simplify session completion logic) We might
close the session before the exchange of the PREPARE_DONE_MESSAGE
message in case 1). To fix, we defer the sending of mutations after
PREPARE_DONE_MESSAGE is sent at the initiator node.
1)
Initiator Follower
tx rx tx rx
1 0 0 1
send prepare
send back prepare
recev prepare
send mutations (close the session before prepare_done msg is sent)
recv mutations (close session before prepare_done msg is received)
send prepare_done
recv prepare_done and send no mutations
2)
Initiator Follower
tx rx tx rx
0 1 1 0
send prepare
send back prepare
recv prepare
nothing to send
send prepare_done
recv prepare_done and send mutations (close session)
recv mutations (close session)
3)
Initiator Follower
tx rx tx rx
1 1 1 1
send prepare
send back prepare
recv prepare
send mutations
recv mutations, can not close session since we have mutations to send
send prepare_done
recv prepare_done and send mutations (close session)
recv mutations (close session)
Message-Id: <d6510b558565db23202164fa491b883ef3796e58.1458634037.git.asias@scylladb.com>
Both the initiator and follower of a stream session knows how many
transfer task and receive task the stream session contains in the
preparation phase. They use the _transfers and _receivers map to track
the tasks, like below:
std::map<UUID, stream_transfer_task> _transfers;
std::map<UUID, stream_receive_task> _receivers;
A stream_transfer_task will send STREAM_MUTATION verb to transfer data
with frozen_mutation, when all the STREAM_MUTATIONs are sent, it will
send STREAM_MUTATION_DONE to tell the peer the stream_transfer_task is
completed and remove the stream_transfer_task from _transfers map. The
peer will remove the corresponding stream_receive_task in _receivers.
We do not really need the COMPLETE_MESSAGE verb to notify the peer we
have completed sending. It makes the session completion logic much
simpler and cleaner if we do not depend on COMPLETE_MESSAGE verb.
However, to be compatible with older version, we always send a
COMPLETE_MESSAGE message and do nothing in the COMPLETE_MESSAGE handler
and replies a ready future even if the stream_session is closed already.
This way, node with older version will get a COMPLETE_MESSAGE message
and manage to send a COMPLETE_MESSAGE message to new node as before.
Message-Id: <1458540564-34277-2-git-send-email-asias@scylladb.com>
For this verb(), we don't call get_session - and it doesn't look like we will.
We currently have no debug message for this one, which makes it harder to debug
the stream of messages. Print it.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Whenever we call get_session, that will print a debug message about the arrival
of this new verb. Because we also print that explicitly in PREPARE_DONE, that
message gets duplicated.
That confuses poor developers who are, for a while, left wondering why is it that
the sender is sender the message twice.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
The cf can be deleted after the cf deletion check. Handle this case as
well.
Use "warn" level to log if cf is missing. Although we can handle the
case, but it is good to distingush where the receiver of streaming
applied all the stream mutations or not. We believe that the cf is
missing because it was dropped, but it could be missing because of a bug
or something we didn't anticipated here.
Related patch: "streaming: Handle cf is deleted when sending
STREAM_MUTATION_DONE"
Fixes simple_add_new_node_while_schema_changes_test failure.
Message-Id: <c4497e0500f50e0a3422efb37e73130765c88c57.1458090598.git.asias@scylladb.com>
Fix bootstrap_test.py:TestBootstrap.failed_bootstap_wiped_node_can_join_test
Logs on node 1:
INFO 2016-03-11 15:53:43,287 [shard 0] gossip - FatClient 127.0.0.2 has been silent for 30000ms, removing from gossip
INFO 2016-03-11 15:53:43,287 [shard 0] stream_session - stream_manager: Close all stream_session with peer = 127.0.0.2 in on_remove
WARN 2016-03-11 15:53:43,498 [shard 0] stream_session - [Stream #4e411ba0-e75e-11e5-81f8-000000000000] stream_transfer_task: Fail to send STREAM_MUTATION_DONE to 127.0.0.2:0: std::runtime_error ([Stream #4e411ba0-e75e-11e5-81f8-000000000000] GOT STREAM_ MUTATION_DONE 127.0.0.1: Can not find stream_manager)
terminate called without an active exception
Backtrace on node 1:
#0 0x00007fb74723da98 in raise () from /lib64/libc.so.6
#1 0x00007fb74723f69a in abort () from /lib64/libc.so.6
#2 0x00007fb74ab84aed in __gnu_cxx::__verbose_terminate_handler() () from /lib64/libstdc++.so.6
#3 0x00007fb74ab82936 in ?? () from /lib64/libstdc++.so.6
#4 0x00007fb74ab82981 in std::terminate() () from /lib64/libstdc++.so.6
#5 0x00007fb74ab82be9 in __cxa_rethrow () from /lib64/libstdc++.so.6
#6 0x0000000000f3521e in streaming::stream_transfer_task::<lambda()>::<lambda(auto:44)>::operator()<std::__exception_ptr::exception_ptr> (ep=..., __closure=0x7ffce74d8630) at streaming/stream_transfer_task.cc:169
#7 do_void_futurize_apply<const streaming::stream_transfer_task::start()::<lambda()>::<lambda(auto:44)>&, std::__exception_ptr::exception_ptr> (func=...) at /home/asias/src/cloudius-systems/scylla/seastar/core/future.hh:1142
#8 futurize<void>::apply<const streaming::stream_transfer_task::start()::<lambda()>::<lambda(auto:44)>&, std::__exception_ptr::exception_ptr> (func=...) at /home/asias/src/cloudius-systems/scylla/seastar/core/future.hh:1190
#9 future<>::<lambda(auto:7&&)>::operator()<future<> > ( fut=fut@entry=<unknown type in /home/asias/src/cloudius-systems/scylla/build/release/scylla, CU 0xec84d00, DIE 0xee2561d>, __closure=__closure@entry=0x7ffce74d8630) at /home/asias/src/cloudius-systems/scylla/seastar/core/future.hh:1014
Message-Id: <1457684884-4776-2-git-send-email-asias@scylladb.com>
In the preparation phase of streaming, we check that remote node has all
the cf_id which are needed for the entire streaming process, including the
cf_id which local node will send to remote node and wise versa.
So, at later time, if the cf_id is missing, it must be that the cf_id is
deleted. It is fine to ingore no_such_column_family exception. In this
patch, we change the code to ignore at server side to avoid sending the
exception back, to avoid handle exception in an IDL compatiable way.
One thing we can improve is that the sender might know the cf is deleted
later than the receiver does. In this case, the sender will send some
more mutations if we send back the no_such_column_family back to the
sender. However, since we do not throw exceptions in the receiver stream
mutation handler, it will not cause a lot of overhead, the receiver will
just ignore the mutation received.
Fixes#979
It is possible that a cf is deleted after we make the cf reader. Avoid
sending them to avoid the unnecessary overhead to send them on the wire and
the peer node to drop the received mutations.
If the peer node of a stream_session is restarted or removed we should
abort the streaming. It is better to hook gossip callback in the stream
manager than in each streamm_session.
When the first time the keep alive timer fires, the _last_stream_bytes
btyes will be zero since it is the first time we update it. The keep
alive timer will be rearmed and fired again. The second time, we find
there is no progress, we close the session. The total idle time will be
2 * keep alive timer.
To make the idle time to close the session be more precise, we reduce
the interval to check the progess and close the session by checking last
time the progress is made.
Message-Id: <c959cffce0cc738a3d73caaf71d2adb709d46863.1456831616.git.asias@scylladb.com>
For each stream_session, we pretend we are sending/receiving one file,
to make it compatible with nodetool. For receiving_files, the file name
is "rxnofile". For sending_files, the file name is "txnofile".
stream_manager::update_all_progress_info is introduced to update the
progress info of all the stream_sessions in the node. We need this
because streaming mutations are received on all the cores, but the
stream_session object is only on one of the cores. It adds overhead if
we update progress info in stream_session object whenever we receive a
streaming mutation. So, what we do now is when we really need the
progress info, we update the progress info in stream_session object.
With http://127.0.0.$i:10000/stream_manager/, it looks like below when
decommission node 3 in a 3 nodes cluster.
=========== GET NODE 1
[{"plan_id": "935a2cc0-dc6b-11e5-bdbf-000000000000", "description":
"Unbootstrap", "sessions": [{"receiving_files": [{"value": {"direction":
"IN", "file_name": "rxnofile", "session_index": 0, "total_bytes":
16876296, "peer": "127.0.0.3", "current_bytes": 16876296}, "key":
"rxnofile"}], "receiving_summaries": [{"files": 1, "total_size": 0,
"cf_id": "869d8630-dc6b-11e5-bdbf-000000000000"}], "session_index": 0,
"state": "PREPARING", "connecting": "127.0.0.3", "peer": "127.0.0.3"}]}]
=========== GET NODE 2
[{"plan_id": "935a2cc0-dc6b-11e5-bdbf-000000000000", "description":
"Unbootstrap", "sessions": [{"receiving_files": [{"value": {"direction":
"IN", "file_name": "rxnofile", "session_index": 0, "total_bytes":
16755552, "peer": "127.0.0.3", "current_bytes": 16755552}, "key":
"rxnofile"}], "receiving_summaries": [{"files": 1, "total_size": 0,
"cf_id": "869d8630-dc6b-11e5-bdbf-000000000000"}], "session_index": 0,
"state": "PREPARING", "connecting": "127.0.0.3", "peer": "127.0.0.3"}]}]
=========== GET NODE 3
[{"plan_id": "935a2cc0-dc6b-11e5-bdbf-000000000000", "description":
"Unbootstrap", "sessions": [{"sending_files": [{"value": {"direction":
"OUT", "file_name": "txnofile", "session_index": 0, "total_bytes":
16876296, "peer": "127.0.0.1", "current_bytes": 16876296}, "key":
"txnofile"}], "sending_summaries": [{"files": 1, "total_size": 0,
"cf_id": "869d8630-dc6b-11e5-bdbf-000000000000"}], "session_index": 0,
"state": "PREPARING", "connecting": "127.0.0.1", "peer":
"127.0.0.1"},{"sending_files": [{"value": {"direction": "OUT",
"file_name": "txnofile", "session_index": 0, "total_bytes": 16755552,
"peer": "127.0.0.2", "current_bytes": 16755552}, "key": "txnofile"}],
"sending_summaries": [{"files": 1, "total_size": 0, "cf_id":
"869d8630-dc6b-11e5-bdbf-000000000000"}], "session_index": 0, "state":
"PREPARING", "connecting": "127.0.0.2", "peer": "127.0.0.2"}]}]
gcc 4.9 complains about the type{ val, val } construction of
type with implicit default constructor, i.e. member = initial
declarations. gcc 5 does not (and possibly rightly so).
However, we still (implicitly) claim to support gcc 4.9 so
why not just change this particular instance.
Message-Id: <1454921328-1106-1-git-send-email-calle@scylladb.com>
Currently, only the shard where the stream_plan is created on will send
streaing mutations. To utilize all the available cores, we can make each
shard send mutations which it is responsbile for. On the receiver side,
we do not forward the mutations to the shard where the stream_session is
created, so that we can avoid unnecessary forwarding.
Note: the downside is that it is now harder to:
1) to track number of bytes sent and received
2) to update the keep alive timer upon receive of the STREAM_MUTATION
To fix, we now store the sent/recieved bytes info on all shards. When
the keep alive timer expires, we check if any progress has been made.
Hopefully, this patch will make the streaming much faster and in turn
make the repair/decommission/adding a node faster.
Refs: https://github.com/scylladb/scylla/issues/849
Tested with decommission/repair dtest.
Message-Id: <96b419ab11b736a297edd54a0b455ffdc2511ac5.1454645370.git.asias@scylladb.com>
The problem is that on the follower side, we set up _session_info too
late, after received PREPARE_DONE_MESSAGE message. The initiator can
send STREAM_MUTATION before sending PREPARE_DONE_MESSAGE message.
To fix, we set up _session_info after we received the prepare_message on
both initiator and follower.
Fixes#869
scylla: streaming/session_info.cc:44: void
streaming::session_info::update_progress(streaming::progress_info):
Assertion `peer == new_progress.peer' failed.
Message-Id: <6d945ba1e8c4fc0949c3f0a72800c9448ba27761.1454476876.git.asias@scylladb.com>
The idea behind the current 10 stream_mutations per core limitation is
to avoid streaming overwhelms the TCP connection and starves normal cql
verbs if the streaming mutations are big and takes long time to
complete.
Now that we use a standalone connection for streaming verbs, we can
increase the limitation.
Hopefully, this will fix#849.
