1. Input and output RowGroup's used in GROUP_CONCAT classes
are currently allocating a raw memory buffer of size equal
to the actual width of the string datatype. As an example,
for the following query:
SELECT col1, GROUP_CONCAT(col2) FROM t GROUP BY col1;
If col2 is a TEXT field with default width, the input
RowGroup containing the target rows to be concatenated will
assign 64kb of memory for every input row in the RowGroup.
This is wasteful as actual field values in real workloads
would be much smaller. We fix this by enabling the
RowGroup to use the StringStore when the RowGroup contains
long strings.
2. RowAggregation::initialize() allocates a memory buffer
for a NULL row. The size of this buffer is equal to the
row size for the output RowGroup. For the above scenario,
using the default group_concat_max_len (which is a server
variable that sets the maximum length of the GROUP_CONCAT string)
value of 1mb, the buffer size would be
(1mb + 64kb + some additional metadata). If the user sets
group_concat_max_len to a higher value, say 3gb, this buffer
size would be ~3gb. Now if the runtime initiates several
instances of RowAggregation, total memory consumption by
PrimProc could exceed the hardware memory limits causing the
OS OOM to kill the process. We fix this problem by again
enabling the StringStore for the NULL row allocation.
3. In the plugin code in buildAggregateColumn(), there is
an integer overflow when the server group_concat_max_len
variable (which is an uint32_t) is set to a value > INT32_MAX
(such as 3gb) and is assigned to
CalpontSystemCatalog::ColType::colWidth (which is an int32_t).
As a short term fix, we saturate the assigned value to colWidth
to INT32_MAX. Proper fix would be to upgrade
CalpontSystemCatalog::ColType::colWidth to an uint32_t.
This patch improves handling of NULLs in textual fields in ColumnStore.
Previously empty strings were considered NULLs and it could be a problem
if data scheme allows for empty strings. It was also one of major
reasons of behavior difference between ColumnStore and other engines in
MariaDB family.
Also, this patch fixes some other bugs and incorrect behavior, for
example, incorrect comparison for "column <= ''" which evaluates to
constant True for all purposes before this patch.
Part 1:
As part of MCOL-3776 to address synchronization issue while accessing
the fTimeZone member of the Func class, mutex locks were added to the
accessor and mutator methods. However, this slows down processing
of TIMESTAMP columns in PrimProc significantly as all threads across
all concurrently running queries would serialize on the mutex. This
is because PrimProc only has a single global object for the functor
class (class derived from Func in utils/funcexp/functor.h) for a given
function name. To fix this problem:
(1) We remove the fTimeZone as a member of the Func derived classes
(hence removing the mutexes) and instead use the fOperationType
member of the FunctionColumn class to propagate the timezone values
down to the individual functor processing functions such as
FunctionColumn::getStrVal(), FunctionColumn::getIntVal(), etc.
(2) To achieve (1), a timezone member is added to the
execplan::CalpontSystemCatalog::ColType class.
Part 2:
Several functors in the Funcexp code call dataconvert::gmtSecToMySQLTime()
and dataconvert::mySQLTimeToGmtSec() functions for conversion between seconds
since unix epoch and broken-down representation. These functions in turn call
the C library function localtime_r() which currently has a known bug of holding
a global lock via a call to __tz_convert. This significantly reduces performance
in multi-threaded applications where multiple threads concurrently call
localtime_r(). More details on the bug:
https://sourceware.org/bugzilla/show_bug.cgi?id=16145
This bug in localtime_r() caused processing of the Functors in PrimProc to
slowdown significantly since a query execution causes Functors code to be
processed in a multi-threaded manner.
As a fix, we remove the calls to localtime_r() from gmtSecToMySQLTime()
and mySQLTimeToGmtSec() by performing the timezone-to-offset conversion
(done in dataconvert::timeZoneToOffset()) during the execution plan
creation in the plugin. Note that localtime_r() is only called when the
time_zone system variable is set to "SYSTEM".
This fix also required changing the timezone type from a std::string to
a long across the system.
