Default timout of 30secs for ConfigRetriever
Default timout of 5sec for use by Transporter (ports etc).
And Ndb_cluster_connection::set_timeout() api for setting timeout from
NDBAPI applications. Should be called before connect. e.g.
c.set_timeout(4200);
c.connect();
Index: ndb-work/storage/ndb/include/mgmcommon/ConfigRetriever.hpp
===================================================================
as side effect - turbo accellerator patch for ndb_mgm - sholud make it quicker... so that 4mhz cpu will seem even faster.
Index: ndb-work/storage/ndb/src/mgmclient/CommandInterpreter.cpp
===================================================================
use portable method, getting milliseconds between calls - Linux would
let us do funky stuff by getting the timeout from select(2). Everywhere
else sucks and doesn't let us do that :(
Index: ndb-work/storage/ndb/include/util/InputStream.hpp
===================================================================
Only one timeout value is needed. Also saves 8bytes per NdbMgmHandle :)
Index: ndb-work/storage/ndb/include/mgmapi/mgmapi.h
===================================================================
Also add an ERROR_codes.txt file for mgmd
Index: ndb-work/storage/ndb/src/mgmsrv/ERROR_codes.txt
===================================================================
In ndb_mgm_call, add checks for expired timeout in (Input|Output)Stream.
In case of timeout, we set NdbMgmHandle->last_error and return NULL.
In api calls not using ndb_mgm_call (or using it in conjunction with
own IO), they'll need to check for timeouts manually. Macros are provided
to do this.
Add ndb_mgm_disconnect_quiet(h) to disconnect without checking errors
(so we don't clobber NdbMgmHandle->last_error). This helps us provide
the *consistent* semantic that on timeout we leave the NdbMgmHandle
*disconnected*. We check for this in testMgm.
Change CHECK_REPLY in mgmapi to also check for set error in handle->last_error
This will pick up the ETIMEDOUT errors and return them to client (through
returning correct failure code for API call and setting NdbMgmHandle error).
Applications written to MGMAPI before this patch will behave as before,
and even hopefully check get_last_error and report the error back to the
end user!
Adding the last CHECK_TIMEDOUT_RET and delete in ndb_mgm_call() we
slightly change behaviour of mgmapi. Previously, if disconnect
midway through a reply, where there were only optional parameters left,
we'd get a Properties object from ndb_mgm_call() containing NULLs for
the optional parameters, leading to interesting error messages. This
enables the returning of the *real* message and actually improves the API
without breaking compatibility.
ndb_mgm_start_signallog
ndb_mgm_stop_signallog
ndb_mgm_log_signals
ndb_mgm_set_trace
ndb_mgm_insert_error
ndb_mgm_set_int64_parameter [1]
ndb_mgm_set_string_parameter [1]
ndb_mgm_purge_stale_sessions [2]
- return error code on error during ndb_mgm_call
TODO:
ndb_mgm_report_event [2]
[1] marked for removal, unused.
[2] return codes incorrect in CHECK_HANDLE/CONNECTED. undocumented.
Server side:
in Services (per session) add macro for injecting timeout error
(just waiting 10 seconds before continuing... it does work!)
We inject these errors in a number of critical places - including
the tricky api functions that don't just use ndb_mgm_call but do
their own thing (get_config, get_status and friends)
ATRT:
Expand testMgm to add timout tests for API. Fully automated.
*THEORETICALLY* timing dependent - an ultra-slow network will
cause problems and "fake" failures... I welcome other solutions.
Tests aren't exhaustive, but cover the generics and the tricky bits.
Also test some calling semantics (incl disconnected on error).
It is encouraged to add *more* mgmapi tests, not less :)
InputStream:
Fix where timedout error is set
Index: ndb-work/storage/ndb/src/mgmapi/mgmapi.cpp
===================================================================
Add not so neat set_timeout call to make testMgm work
Index: ndb-work/storage/ndb/include/mgmapi/mgmapi.h
===================================================================
Add error injection either for this connection or for whole server.
Currently nothing for injecting errors into *another* connection... but that's
perhaps getting tricky-dicky for this point in time. Perhaps needed for events
if we don't do anything fancy.
Index: ndb-work/storage/ndb/src/mgmsrv/MgmtSrvr.cpp
===================================================================
Start using the write_timeout we already have in NdbMgmHandle
Index: ndb-work/storage/ndb/src/mgmapi/mgmapi.cpp
===================================================================
The problem in this bug is when we create temporary tables. When
temporary tables are created for unions, there is some
inferrence being carried out regarding the type of the column.
Whenever this column type is inferred to be REAL (i.e. FLOAT or
DOUBLE), MySQL will always try to maintain exact precision, and
if that is not possible (there are hardware limits, since FLOAT
and DOUBLE are stored as approximate values) will switch to
using approximate values. The problem here is that at this point
the information about number of significant digits is not
available. Furthermore, the number of significant digits should
be increased for the AVG function, however, this was not properly
handled. There are 4 parts to the problem:
#1: DOUBLE and FLOAT fields don't display their proper display
lengths in max_display_length(). This is hard-coded as 53 for
DOUBLE and 24 for FLOAT. Now changed to instead return the
field_length.
#2: Type holders for temporary tables do not preserve the
max_length of the Item's from which they are created, and is
instead reverted to the 53 and 24 from above. This causes
*all* fields to get non-fixed significant digits.
#3: AVG function does not update max_length (display length)
when updating number of decimals.
#4: The function that switches to non-fixed number of
significant digits should use DBL_DIG + 2 or FLT_DIG + 2 as
cut-off values (Since fixed precision does not use the 'e'
notation)
Of these points, #1 is the controversial one, but this
change is preferred and has been cleared with Monty. The
function causes quite a few unit tests to blow up and they had
to b changed, but each one is annotated and motivated. We
frequently see the magical 53 and 24 give way to more relevant
numbers.
fix for cast( AS DATETIME) + 0 operation.
I just implemented Item_datetime_typecast::val() method
as it is usually done in other classes.
Should be fixed more radically in 5.0
of its argument happened to be a decimal expression returning
the NULL value.
The crash was due to the fact the function in_decimal::set did
not take into account that val_decimal() could return 0 if
the decimal expression had been evaluated to NULL.
