Given the fact that there exist only three possible categories, the map
can be replaced with a static array that needs no memory
allocations. Making this array thread-local allows it to be reused which
places an upper limit on the number of memory allocations.
The documentation stated that at most `max_sescmd_history` commands were
kept but in reality the number of commands kept in the history was one
command smaller than what was documented.
This commit adds a new parameter that, when enabled, prunes the session
command history to a known length. This makes it possible to keep a
client-side pooled connection open indefinitely at the cost of making
reconnections theoretically unsafe. In practice the maximum history length
can be set to a value that encompasses a single session using the pooled
connection with no risk to session state integrity. The default history
length of 50 commands is quite likely to be adequate for the majority of
use-cases.
When the connection state is reset by executing a COM_CHANGE_USER or
COM_RESET_CONNECTION, readwritesplit does not need to store the session
command history that was executed before it. With this, pooled connections
will effectively behave like normal connections if the pooling mechanism
is smart enough to reset the connection. This also prevents unwanted
visibility into the session states of other connections.
If the routing of a session command fails due to problems with the backend
connections, a more verbose error message is logged. The added status
information in the Backend class makes tracking the original cause of the
problem a lot easier due to knowing where, when and why the connection was
closed.
If a server was not chosen as the target of a routing hint, the server
status would not be logged. By logging the server state in the message, it
is easier to figure out why a server wasn't chosen as the routing target.
Both the replication lag and the message printing state are saved in SERVER,
although the values are mostly used by readwritesplit. A log message is printed
both when a server goes over the limit and when it comes back below.
Because of concurrency issues, a message may be printed multiple times before
different threads detect the new message state.
Documentation updated to explain the change.
If the connection to the master is lost, knowing what type of an error
caused the call to handleError helps deduce what was the real reason for
it. Logging the idle time of the connection helps detect when the
wait_timeout of a connection is exceeded.
By storing the server statistics object in side the session, the lookup
involved in getting a worker-local value is avoided. Since the lookup is
done multiple times for a single query, it is beneficial to store it in
the session.
As the worker-local value is never deleted, it is safe to store a
reference to it in the session. It is also never updated concurrently so
no atomic operations are necessary.
The code now only checks the need for a keepalive ping once every
keepalive interval. Reduced the number of mxs_clock calls to one so that
all servers use the same value.
The information stored for each prepared statement would not be cleared
until the end of the session. This is a problem if the sessions last for a
very long time as the stored information is unused once a COM_STMT_CLOSE
has been received.
In addition to this, the session command response maps were not cleared
correctly if all backends had processed all session commands.
By resetting the replay state the transaction replay can start again on a
new server. This allows the replay process work when a master server is
shutting down.
By delaying the replay for a second, we give the monitor a small chance to
adap to master failures. It'll also prevent rapid re-querying if multiple
transaction replays are supported.
A transaction that just completed will go through the start_trx_replay
function as from the client protocol's point of view the transaction is
still open. The debug assertion did not take this into account and would
fail if a successful commit was the last thing done on master that failed.
Also fixed the formatting.
When a server is stopping, it'll send an error to the client before
terminating the TCP connection. The code in readwritesplit would detect
this error and create a hangup event on the DCB. This would cause it to
appear as if the TCP connection was broken and the router would
immediately try to reconnect to the same server.
By ignoring the error and allowing the connection to die on its own, we
avoid immediately reconnecting and retrying any transactions on the
stopping server. This increases the chances that the monitor will see it
first and assign the server states correctly before the transaction replay
is attempted.
The assertion would hold true for a single worker but it can't be
guaranteed to be true on a multi-worker system where the statistics are
distributed across the workers.
Enabling the feature by default prevents the master connection from dying
during times when there are very little or no writes. Having a modest ping
interval of 300 seconds serves to minimize the amount of extra work that
both MaxScale and the server have to do while still keeping the
connections in good shape.
If the server where a query is being executed is shutting down,
readwritesplit should treat it as an error to make retrying of the query
possible.
By treating server shutdowns as network errors, the same code path that is
used for actual network errors can be taken. This removes the need for any
extra retrying logic for this particular case.
If the master succeeds in executing a session command but the slave fails,
the error message could help explain why it failed. At the moment this is
mainly relevant for inspection of test results.
The transaction replay could get mixed up with new queries if the client
managed to perform one while the delayed routing was taking place. A
proper way to solve this would be to cork the client DCB until the
transaction is fully replayed. As this change would be relatively more
complex compared to simply labeling queries that are being retried the
corking implementation is left for later when a more complete solution can
be designed.
This commit also adds some of the missing info logging for the transaction
replaying which makes analysis of failures easier.
The servers with a zero weight would be always used over ones that have a
weight. This means that the behavior was inverted and caused the
mxs2054_hybrid_cluster test to fail in 2.3.
Also fixed a typo in the deprecation message.
Commit a9e236497963251f8b4afa07484b88ad97e73a03 changed where the PS ID
for a binary protocol command is replaced with the internal form. This
caused prepared statements that are also session commands to be always
routed with the external ID.
As the external ID is almost always the master's ID, the aforementioned
bug resulted in odd side-effects and the true cause of these was only
revealed when the error message sent by the slave was included in the log
messages.
If a PS command is routed multiple times, the ID will not be reverted to
the external ID in the failure cases. This prevented prepared statements
from being re-routed correctly.
When the connection to the master is broken, the session is not configured
to use the read-only modes and the monitor can still connect to the
server, the connection will be closed and and error is sent to the
client. To leave some trace of this problem in the MaxScale logs, a
message should always be logged when a network error occurs.
As the router is the only one that knows what backends a particular
statement has been sent to, it is the responsibility of the router
to keep the session bookkeeping up to date. If it doesn't we will
know what statements a session has received (provided at least some
component in the routing chain has RCAP_TYPE_STMT_INPUT capability),
but not how long their processing took. Currently only readwritesplit
does that.
All queries are stored and not just COM_QUERY as that makes the
overall bookkeeping simpler; at clientReply() time we do not need to
know whether or not to bookkeep information, we can just do it.
When session information is queried for, we report as much information
we have available.
The causal_reads_timeout default value is too long when considering the
behavioral changes that MXS-2141 introduced. With a 10 second default
value, a result is returned to the client in a reasonable amount of time.
With causal_reads enabled, the query would return with an error if the
slave was not able to catch up to the master fast enough. By automatically
retrying the query on the master, we're guaranteed that a valid result is
always returned to the client.
