Ada Böhm

* Some changes are summarized in docs and changelog * Benchmarks are not ported yet * ~~PyHQ is not ported yet.~~

When task is canceled (e.g. via "hq job cancel ") then it does not kill its child processes. It is a problem when the user starts a bash script as...

Allow to define task requests with multi resource requests: e.g: [1 cpu + 1 gpu **OR** 16 cpus] Usecase: Let us have a node with a few GPUs and many...

* [x] Basic functionality * [x] Multi-node task parking * [ ] Switch for requesting only workers from the same allocation * [ ] Dynamic requests of number-of-nodes-per-allocation from auto...

Storing worker monitoring into binary log

2

* Add new type of block into binary stream (file_writer in stream/server/rpc.rs) that contains serialized JSON with monitoring data. * Modify reader (stream/reader/logfile.rs) tthat it is still able read stream

Autoalloc improvements

5

Improve the existing implementation of autoallocation. - [x] #128 - [x] #129 - [x] #145 - [ ] Match HQ workers with allocations. Currently, we have no information about which...

Auto allocation table size

1

`hq alloc info ` produces large tables. It cannot be snown in my console. Some shortening? More lines per entry?

Better task sandboxing

3

When task spawn more processes, make sure that they are all killed when task is canceled or finished.

Implement * [x] ``hq worker stop `` - stop worker right now * [ ] ``hq worker stop --soft`` - stop schedule tasks into this worker, when all tasks on...

- ~~Generic resource management~~ - Persistent information about jobs