SO_REUSEPORT is a socket option that allows multiple servers to listen
on the same TCP port, and the kernel will automatically load balance the
connections between them. This is particularly helpful for the hash
server since it runs in a single thread. To take advantage of a
multi-core server, multiple servers can be started in parallel with this
option (up to 1 per CPU) and the kernel will load balance between them.
(Bitbake rev: d72d5a7decb489e2af0ebc43cfea0ca3e4353e9b)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Adds support for asyncrpc servers to send connection headers to clients
on connection. Since this is a breaking protocol change, clients must
opt-in to expect headers from the server, corresponding to a version
bump in the client protocol.
(Bitbake rev: 1cb2b8be6cc5269553f549285592e47b7d29db03)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When running an asyncrpc server as a subprocess, it is often desired to
run it with a lower logging level since the normal logging of clients
connecting and disconnecting is not desired.
As such, add an option to set the logging level of the server when
running as a subprocess and set the level to WARNING when starting a
local hashserver or PRserver
(Bitbake rev: 61dac7b99ad6d2a858f85d8ed1b5524d558be6c8)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Adds support for Invocation Errors (that is, errors raised by the actual
RPC call instead of at the protocol level) to propagate across the
connection. If a server RPC call raises an InvokeError, it will be sent
across the connection and then raised on the client side also. The
connection is still terminated on this error.
(Bitbake rev: 50ee68175e7cf20a32bfbb176db2c47d7859da04)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Adds a logging adaptor to the asyncrpc clients that prefixes log
messages with the client remote address to aid in debugging
(Bitbake rev: f4d64ce73c2449c008ff5d9b32376a2893ef7195)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Adds support to the hash equivalence client and server to communicate
over websockets. Since websockets are message orientated instead of
stream orientated, and new connection class is needed to handle them.
Note that websocket support does require the 3rd party websockets python
module be installed on the host, but it should not be required unless
websockets are actually being used.
(Bitbake rev: 56dd2fdbfb6350a9eef43a12aa529c8637887a7e)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Rewrites the asyncrpc client and server code to make it possible to have
other transport backends that are not stream based (e.g. websockets
which are message based). The connection handling classes are now shared
between both the client and server to make it easier to implement new
transport mechanisms
(Bitbake rev: 2aaeae53696e4c2f13a169830c3b7089cbad6eca)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
If the client socket is closed, asyncio.StreamReader.readline() will
return an empty bytes object, not None.
This prevents multiple tracebacks being logged by bitbake-hashserv each
time bitbake is started and performs a connection check.
(Bitbake rev: 2d07f252704dff7747fa1f9adf223a452806717f)
Signed-off-by: Justin Bronder <jsbronder@cold-front.org>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Where copyright headers were not present, add them to make things
clear.
(Bitbake rev: 1aa338a216350a2751fff52f866039343e9ac013)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Adds TCP Keep Alive support to the async RPC server. This should help
prevent file descriptor exhaustion on the server when client connections
are interrupted and the socket never closes (e.g. no FIN is sent from
the client).
A keep alive is sent after 30 seconds of inactivity, then every 15
seconds after that up to a maximum of 2 minutes.
(Bitbake rev: 68f4ce662cad28fed739900addbdee949ad3c1e8)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
asyncio in older Python 3.x (seen with 3.7) can seemingly hang if
new_event_loop is called repeatedly in the same process. The
reuse of processes in the Bitbake thread pool during parsing seems
to be able to trigger this with the PR server export selftest.
It appears that calling set_event_loop with the new loop avoids the
issue, so that is now done in the asyncrpc Client initializer (with
an explanatory comment). This should be revisited when the day
arrives that Python 3.9 becomes the minimum required for BitBake.
Additionally, it was discovered that using get_event_loop in the
asyncrpc server initialization can trigger hangs in the hashserv
unittests when the second test is run. To avoid this, switch to
calling new_event_loop + set_event_loop in the initialization code
there as well.
(Bitbake rev: bb9a36563505652294b20b4c88199b24fa208342)
Signed-off-by: Scott Murray <scott.murray@konsulko.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Reworks the async I/O API so that the async loop is only created in the
child process. This requires deferring the creation of the server until
the child process and a queue to transfer the bound address back to the
parent process
(Bitbake rev: 8555869cde39f9e9a9ced5a3e5788209640f6d50)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
[small loop -> self.loop fix in serv.py]
Signed-off-by: Scott Murray <scott.murray@konsulko.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
If the SIGTERM signal is sent to an asyncrpc server before it has
installed the SIGTERM handler in the main loop, it may miss the signal
which will can cause the calling process to wait forever on the join().
To resolve this, the calling process should mask of SIGTERM before
forking the server process and the server should unmask the signal only
after the handler is installed. To simplify the usage of the server, an
new helper function called serve_as_process() is added to do this
automatically and correctly.
Thanks: Scott Murray <scott.murray@konsulko.com> for helping debug
(Bitbake rev: ef2865efa98ad20823267364f2159d8d8c931400)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This method is needed to support startup of the prservice. As it is so
generic we can add it to the common asyncrpc module.
(Bitbake rev: 25ccd697ea76f66b813be2296866b2d3405b079c)
Signed-off-by: Paul Barker <pbarker@konsulko.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The hashserv module implements a flexible RPC mechanism based on sending
json formatted messages over unix or tcp sockets and uses Python's
asyncio features to build an efficient message loop on both the client
and server side. Much of this implementation is not specific to the
hash equivalency service and can be extracted into a new module for
easy re-use elsewhere in bitbake.
(Bitbake rev: 4105ffd967fa86154ad67366aaf0f898abf78d14)
Signed-off-by: Paul Barker <pbarker@konsulko.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
