Fix load track behavior on multiple load driver machines#1315
Fix load track behavior on multiple load driver machines#1315DJRickyB merged 20 commits intoelastic:masterfrom
Conversation
This reverts commit 39ae4f3.
DJRickyB
added
bug
|
Further testing seems to indicate some silent hanging, I will circle back to this on Monday. Feedback welcome on current state |
dliappis
left a comment
dliappis
left a comment
There was a problem hiding this comment.
Thank you fixing this! I did an initial pass, looking only at the code, which looks good; left a few minor comments.
Will test a few scenarios, as the next step.
| def __init__(self, worker_id, config, track, client_allocations): | ||
| """ | ||
| :param worker_id: Unique (numeric) id of the worker. | ||
| :param config: Rally internal configuration object. |
There was a problem hiding this comment.
Was this removed on purpose?
There was a problem hiding this comment.
whoops! adding it back. I had removed config from this message altogether in an earlier iteration
esrally/driver/driver.py
Outdated
|
|
||
| class RallyConfig: | ||
| """ | ||
| Prompts loading of a track |
There was a problem hiding this comment.
It's a bit unclear to me what this is actually doing, by just reading the docstring. I fail to understand how an object instantiating this that stores cfg prompts (as in causes/brings about?) the loading of a track. Is there anything short we could add here to make this easier to understand in the future without having to git blame and reference the issue #1206?
There was a problem hiding this comment.
Great point. I think I will go the "self-documenting" route here. Other message classes are verbs, telling the actor what to do, this one would be better named "Bootstrap" and could in the future denote other activity actors should be doing before their intended work.
dliappis
left a comment
dliappis
left a comment
There was a problem hiding this comment.
I've now concluded tests, things seem to be working fine with large loaddriver clusters.
Also the latest commits after my previous comments are great.
LGTM.
2 participants
This commit fixes #1206 by causing the TrackPreparationActor and the Driver actor to register the local track repository/code prior to attempting to receive the loaded track as a serialized message.