Fix error in "extractAllGroups" function

[alexey-milovidov]

Changelog category (leave one):

• Bug Fix

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):
Fix the issue when some invocations of extractAllGroups function may trigger "Memory limit exceeded" error. This fixes #13383

[alesapin]

AST Fuzzer:

 56 Aborted                 (core dumped) ./clickhouse-server --config-file db/config.xml -- --path db 2>&1

But no errors in the server log and in the fuzzer log.

[akuzm]

AST Fuzzer:

 56 Aborted                 (core dumped) ./clickhouse-server --config-file db/config.xml -- --path db 2>&1

But no errors in the server log and in the fuzzer log.

No errors would be bad, better investigate. But I found the error in server log:

2020.09.17 03:44:13.654828 [ 71 ] {} <Trace> SystemLog (system.trace_log): Flushing system log, 24 entries to flush
<jemalloc>: ../contrib/jemalloc/include/jemalloc/internal/arena_inlines_a.h:45: Failed assertion: "newarena != NULL"
2020.09.17 03:44:13.657126 [ 63 ] {} <Trace> BaseDaemon: Received signal 6

It also looks very bad...

[akuzm]

It also looks very bad...

1. jemalloc shouldn't have an assertion in this place -- a creation of new arena can return null in case we are out of memory
2. there are a lot of messages like these:

It looks like that the process is near the limit on number of virtual memory mappings.
Current number of mappings (/proc/self/maps): 1048566.
Limit on number of mappings (/proc/sys/vm/max_map_count): 1048576.
You should increase the limit for vm.max_map_count in /etc/sysctl.conf

But why would we be out of memory mappings on a server that is basically idle? Looks like a memory leak.