- Progress bar (and query estimation, requires driver support)
- Diff profile events for multiple queries
ProfileEvents
in a loadavg fashion (1/5/15 usingsimple_moving_average
crate)- Configurable columns
- New metrics (page cache usage, but there are some issues with this metrics in ClickHouse itself)
- Colored queries metrics (if uses too much RAM/CPU/Disk/Net)
- Graphs for summary metrics (memory, ...)
- Compare multiple queries (
ProfileEvents
) system.trace_log
->system.stack_trace
(bythread_id
)- implement
system.kernel_stack_trace
and support it here - look at how much does it spent time in locks (but care should be take and conditional variables should not be take into account)
- various grouping
- implement
- Decompose query to the inner most subquery
- Async metrics with charts
See lot's of TODO/FIXME/NOTE in the code
system.events/metrics/asynchronous_metrics
system.parts
system.warnings
system.*_log
system.processors_profile_log JOIN system.query_log
- ...
- Locks introspection
- ...
- better Rust
- better error handling
- better shortcuts
- write tests
- extend documentation (Features, Motivation)
- add screencasts with asciinema
- Rewrite flameshow in Rust (at least brew does not accept packages that depends on PyOxidizer)
- panic from thread fails only that thread, it need to stop the whole program