Fix buffer size for trace collection

[azat]

Changelog category (leave one):

• Not for changelog (changelog entry is not required)

[alexey-milovidov]

BTW query profiler can profile itself (for example Real time profiler can measure how frequently CPU time profiler was active). This is somewhat useful property.

[azat]

BTW query profiler can profile itself (for example Real time profiler can measure how frequently CPU time profiler was active). This is somewhat useful property.

I see, thanks. Okay, let's truncate the buffer size to 4K then.

[alexey-milovidov]

#21137

[alexey-milovidov]

@azat Do you know why fuzzer did not stop after first segfault?

[azat]

@azat Do you know why fuzzer did not stop after first segfault?

It should be because SIGSEGV handler does not call _exit, so when thread got SIGSEGV it just send information about the signal to the pipe, and the server will be terminated only once this message in pipe will be served, but this may take a while, in the meantime SIGSEGV in another handler may happen.

[alexey-milovidov]

But they are from different queries (check query_id) and fuzzer is sending queries sequentially.

[azat]

But they are from different queries (check query_id) and fuzzer is sending queries sequentially.

Yep, but the problem is that the first query failed with different error - Data compressed with different methods:

Error on processing query 'SELECT item_id FROM (SELECT item_id FROM t GROUP BY item_id WITH TOTALS) AS l FULL OUTER JOIN (SELECT item_id FROM t GROUP BY item_id WITH TOTALS) AS r USING (item_id)': Code: 271, e.displayText() = DB::Exception: Data compressed with different methods, given method byte 0x6e, previous method byte 0x82: while receiving packet from localhost:9000, Stack trace (when copying this message, always include the lines below):
Error on processing query 'SELECT item_id FROM (SELECT item_id FROM t GROUP BY item_id WITH TOTALS) AS l FULL OUTER JOIN (SELECT substring(randomString(0), NULL), item_id FROM t GROUP BY item_id WITH TOTALS) AS r USING (item_id)': Code: 271, e.displayText() = DB::Exception: Data compressed with different methods, given method byte 0x6e, previous method byte 0x82: while receiving packet from localhost:9000, Stack trace (when copying this message, always include the lines below):
Error on processing query 'SELECT item_id FROM (SELECT item_id FROM t GROUP BY item_id WITH TOTALS) AS l FULL OUTER JOIN (SELECT arrayMap(x -> reinterpretAsUInt8(substring(2, randomString(NULL), x + 65536)), range(1048577)), item_id FROM t GROUP BY item_id WITH TOTALS) AS r USING (item_id)': Code: 271, e.displayText() = DB::Exception: Data compressed with different methods, given method byte 0x6e, previous method byte 0x82: while receiving packet from localhost:9000, Stack trace (when copying this message, always include the lines below):
Error on processing query 'SELECT item_id FROM (SELECT item_id FROM t GROUP BY arrayJoin(arrayMap(x -> reinterpretAsUInt8(substring(randomString(1048576), x + 256, NULL)), range(256))), item_id WITH TOTALS) AS l FULL OUTER JOIN (SELECT item_id FROM t GROUP BY item_id WITH TOTALS) AS r USING (item_id)': Code: 271, e.displayText() = DB::Exception: Data compressed with different methods, given method byte 0x6e, previous method byte 0x82: while receiving packet from localhost:9000, Stack trace (when copying this message, always include the lines below):

This is because fatal thread handler was called when the TCPHandler was sending totals:

4  0x00007ffff7671909 in sleepForSeconds (seconds=20) at sleep.cpp:64
5  0x00007ffff7e97674 in signalHandler (sig=6, info=0x7fff4effecf0, context=0x7fff4effebc0) at BaseDaemon.cpp:155
6  <signal handler called>
7  0x00007ffff5fb5615 in raise () from /usr/lib/libc.so.6
8  0x00007ffff5f9e862 in abort () from /usr/lib/libc.so.6
9  0x00007ffff796b182 in DB::handle_error_code (msg="Bad cast from type DB::ColumnVector<char8_t> to DB::ColumnNullable", code=49) at Exception.cpp:47
10 0x00007ffff796b26e in DB::Exception::Exception (this=0x7fff415d9c80, msg="Bad cast from type DB::ColumnVector<char8_t> to DB::ColumnNullable", code=49, remote_=false) at Exception.cpp:57
11 0x00007ffff3639f3f in assert_cast<DB::ColumnNullable const&, DB::IColumn const&> (from=...) at assert_cast.h:47
12 0x00007fffe270edbc in DB::DataTypeNullable::serializeBinaryBulkWithMultipleStreamsImpl () at DataTypeNullable.cpp:92
13 0x00007fffe2779680 in DB::IDataType::serializeBinaryBulkWithMultipleStreams () at IDataType.cpp:286
14 0x00007fffe2af2485 in DB::NativeBlockOutputStream::writeData () at NativeBlockOutputStream.cpp:58
15 0x00007fffe2af2956 in DB::NativeBlockOutputStream::write () at NativeBlockOutputStream.cpp:124
16 0x00007fffdcabc820 in DB::TCPHandler::sendTotals () at TCPHandler.cpp:763

And client will have 20 seconds to send other queries