Make the execmanager.retrieve_calculation idempotent'ish
#3142
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The `retrieve_calculation` would cause an exception if called multiple times for the same calculations, which can happen if the first time that the runner was working on it got interrupted, for example due to a daemon shutdown. The reason is that the second time around the adding of the `retrieved` folder data node will raise a uniqueness exception, because there can only be one output with the same label. Note that full idempotency is impossible, but this change should make the problem a lot less likely to occurr. The idea is to delay the actual attaching of the retrieved folder data node to the last moment possible. This way, if the method is called again and the folder is already there, we can be reasonably sure that the files were already retrieved succesfully and we simply return, leaving the call a no-op. This is done in the beginning of the function to check if the output node already exists using the `LinkManager.first()` call. If the node exists, the retrieve function has apparently already been called before and reached the end of the function where it adds the retrieved folder. This means all the files were already successfully retrieved so we can safely skip. The `LinkManager.first()` method is adapted to, instead of raising a `ValueError` when no entry is found at all, simply return `None`. This is more consistent with the behavior of `QueryBuilder.first()`.
sphuber
force-pushed
the
fix_3141_execmanager_retrieve_calculation_idempotence
branch
from
19e3e84 to
81de7b2
Compare
giovannipizzi
approved these changes
Jul 9, 2019
There was a problem hiding this comment.
Ok - ideally, the final solution would be to make sure the add_incoming an atomic operation (wrapping it in a transaction?). I think we can approve this, maybe can you add a note to check/improve this in the issue where we should address atomicity, transactions etc?
I also agree with returning None in first(), this is consistence with SQLAlchemy's first() method vs. one() that instead raises
d-tomerini
pushed a commit
to d-tomerini/aiida_core
that referenced
this pull request
Sep 30, 2019
…#3142) The `retrieve_calculation` would cause an exception if called multiple times for the same calculations, which can happen if the first time that the runner was working on it got interrupted, for example due to a daemon shutdown. The reason is that the second time around the adding of the `retrieved` folder data node will raise a uniqueness exception, because there can only be one output with the same label. Note that full idempotency is impossible, but this change should make the problem a lot less likely to occur. The idea is to delay the actual attaching of the retrieved folder data node to the last moment possible. This way, if the method is called again and the folder is already there, we can be reasonably sure that the files were already retrieved successfully and we simply return, leaving the call a no-op. This is done in the beginning of the function to check if the output node already exists using the `LinkManager.first()` call. If the node exists, the retrieve function has apparently already been called before and reached the end of the function where it adds the retrieved folder. This means all the files were already successfully retrieved so we can safely skip. The `LinkManager.first()` method is adapted to, instead of raising a `ValueError` when no entry is found at all, simply return `None`. This is more consistent with the behavior of `QueryBuilder.first()`.
d-tomerini
pushed a commit
to d-tomerini/aiida_core
that referenced
this pull request
Oct 16, 2019
…#3142) The `retrieve_calculation` would cause an exception if called multiple times for the same calculations, which can happen if the first time that the runner was working on it got interrupted, for example due to a daemon shutdown. The reason is that the second time around the adding of the `retrieved` folder data node will raise a uniqueness exception, because there can only be one output with the same label. Note that full idempotency is impossible, but this change should make the problem a lot less likely to occur. The idea is to delay the actual attaching of the retrieved folder data node to the last moment possible. This way, if the method is called again and the folder is already there, we can be reasonably sure that the files were already retrieved successfully and we simply return, leaving the call a no-op. This is done in the beginning of the function to check if the output node already exists using the `LinkManager.first()` call. If the node exists, the retrieve function has apparently already been called before and reached the end of the function where it adds the retrieved folder. This means all the files were already successfully retrieved so we can safely skip. The `LinkManager.first()` method is adapted to, instead of raising a `ValueError` when no entry is found at all, simply return `None`. This is more consistent with the behavior of `QueryBuilder.first()`.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #3141
The
retrieve_calculationwould cause an exception if called multipletimes for the same calculations, which can happen if the first time that
the runner was working on it got interrupted, for example due to a
daemon shutdown. The reason is that the second time around the adding of
the
retrievedfolder data node will raise a uniqueness exception,because there can only be one output with the same label.
Note that full idem-potency is impossible, but this change should make
the problem a lot less likely to occur. The idea is to delay the actual
attaching of the retrieved folder data node to the last moment possible.
This way, if the method is called again and the folder is already there,
we can be reasonably sure that the files were already retrieved
successfully and we simply return, leaving the call a no-op. This is done
in the beginning of the function to check if the output node already
exists using the
LinkManager.first()call.The
LinkManager.first()method is adapted to, instead of raising aValueErrorwhen no entry is found at all, simply returnNone. Thisis more consistent with the behavior of
QueryBuilder.first().