"After the patch, all of our relevant I/O is placed on a specific priority class.
The ones which are not are left into the Seastar's default priority, which will
effectively work as an idle class.
Examples of such I/O are commitlog replay and initial SSTable loading. Since they
will happen during initialization, they will run uncontended, and do not justify
having a class on their own."
After this patch, our I/O operations will be tagged into a specific priority class.
The available classes are 5, and were defined in the previous patch:
1) memtable flush
2) commitlog writes
3) streaming mutation
4) SSTable compaction
5) CQL query
Signed-off-by: Glauber Costa <glauber@scylladb.com>
After the introduction of the Fair I/O Queueing mechanism in Seastar,
it is possible to add requests to a specific priority class, that will
end up being serviced fairly.
This patch introduces a Priority Manager service, that manages the priority
each class of request will get. At this moment, having a class for that may
sound like an overkill. However, the most interesting feature of the Fair I/O
queue comes from being able to adjust the priorities dynamically as workloads
changes: so we will benefit from having them all in the same place.
This is designed to behave like one of our services, with the exception that
it won't use the distributed interface. This is mainly because there is no
reason to introduce that complexity at this point - since we can do thread local
registration as we have been doing in Seastar, and because that would require us
to change most of our tests to start a new service.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
SSTables already have a priority argument wired to their read path. However,
most of our reads do not call that interface directly, but employ the services
of a mutation reader instead.
Some of those readers will be used to read through a mutation_source, and those
have to patched as well.
Right now, whenever we need to pass a class, we pass Seastar's default priority
class.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
All the SSTable read path can now take an io_priority. The public functions will
take a default parameter which is Seastar's default priority.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
All variants of write_component now take an io_priority. The public
interfaces are by default set to Seastar's default priority.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
The only user for the default size is data_read, sitting at row.cc.
That reader wants to read and process a chunk all at once. So there's
really no reason to use the default buffer size - except that this code
is old.
We should do as we do in other single-key / single-range readers and
try to read all at once if possible, by looking at the size we received
as a parameter. Cleaning up the data_stream_at interface then comes as
a nice side effect.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Its definition as a lambda function is inconvenient, because it does not allow
us to use default values for parameters.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Its definition as a lambda function is inconvenient, because it does not allow
us to use default values for parameters.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
In the existing code, when we fail to reach one of the replicas of some
range being repaired, we would give up, and not continue to repair the
living replicas of this range. The thinking behind this was since the
repair should be considered failed anyway, there's no point in trying
to do a half-job better.
However, in a discussion I had with Shlomi, he raised the following
alternative thinking, which convinced me: In a large cluster, having
one node or another temporarily dead has a high probability. In that
case, even if the if the repair is doomed to be considered "failed",
we want it at least to do as much as it possibly can to repair the
data on the living part of the cluster. This is what this patch does:
If we can only reach some of the replicas of a given range, the repair
will be considered failed (as before), but we will still repair the
reachable replicas of this range, if they have different checksums.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <1453724443-29320-1-git-send-email-nyh@scylladb.com>
Unlike streaming in c*, scylla does not need to open tcp connections in
streaming service for both incoming and outgoing messages, seastar::rpc
does the work. There is no need for a standalone stream_init_message
message in the streaming negotiation stage, we can merge the
stream_init_message into stream_prepare_message.
The helper is used only once in init_sending_side and in
init_receiving_side we do not use create_and_register to create
stream_result_future. Kill the trivial helper to make the code more
consistent.
In addition, rename variables "future" and "f" to sr (streaming_result).
So we have:
- init_sending_side
called when the node initiates a stream_session
- init_receiving_side
called when the node is a receiver of a stream_session initiated by a peer
In scylla, in each stream_coordinator, there will be only one
stream_session for each remote peer. Drop the code supporting multiple
stream_sessions in host_streaming_data.
We now have
shared_ptr<stream_session> _stream_session
instead of
std::map<int, shared_ptr<stream_session>> _stream_sessions
- from
We can get it form the rpc::client_info
- session_index
There will always be one session in stream_coordinator::host_streaming_data with a peer.
- is_for_outgoing
In cassandra, it initiates two tcp connections, one for incoming stream and one for outgoing stream.
logger.debug("[Stream #{}] Sending stream init for incoming stream", session.planId());
logger.debug("[Stream #{}] Sending stream init for outgoing stream", session.planId());
In scylla, it only initiates one "connection" for sending, the peer initiates another "connection" for receiving.
So, is_for_outgoing will also be true in scylla, we can drop it.
- keep_ss_table_level
In scylla, again, we stream mutations instead of sstable file. It is
not relevant to us.
