* use AggregationMethod ctor with reserve
* add new settings
* add HashTablesStatistics
* support queries with limit
* support distributed and with external aggregation
* add new profile events
* add some tests
* add perf test
* export cache stats through AsynchronousMetrics
* rm redundant trace
* fix style
* fix 02122_parallel_formatting test
* review fixes
* fix 02122_parallel_formatting test
* apply also to two-level HTs
* try simpler strategy
* increase max_size_to_preallocate_for_aggregation for experiment
* fixes
* Revert "increase max_size_to_preallocate_for_aggregation for experiment"
This reverts commit 6cf6f75704.
* fix test
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This allows starting and stopping separately each protocol server
without restarting ClickHouse.
This also allows adding or removing `listen_host` entries, which
start and stops servers for all enabled ports.
When stopping a server, the listening socket is immediately closed
(and available for another server).
Protocols with persistent connections try to wait for any currently
running query to finish before closing the connection, but idle
connection are closed quickly (depending on how often the protocol
is polled).
An extra ProfileEvent is added, `MainConfigLoads`, it is
incremented every time the configuration is reloaded. This helps
when trying to assess whether the new configuration was applied.
- rescan /sys/block on errors (disk had been replaced and you will got
ENOENT)
- rescan /sys/block periodically (each 5m), for newly added devices
(i.e. new disk to RAID array can be added)
