Let libcurl downloads be managed via lockfiles #3735
Merged
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
We want to avoid race conditions on gmt_data_server.txt and gmt_hash_server.txt.
If for whatever reason the gmt_data_server.txt ends up blank, reading will fail. Rather than just complain about it, we now delete the file since we (a) know it is broken and (b) unless it is removed we will not attempt to refresh it for another cycle (by default 24 hours).
|
The latest commit (delete gmt_server_data.txt if faulty) should also help as it will allow a refresh teh next time accessed. Note that the gmtserver apparently has powered down due to the hurricane so you cannot really test on oceania until we resurface. |
PaulWessel
changed the title
|
THis works well for me and I have not had any gmt_data_server.txt errors since adding the flock and remove. |
This was referenced Jul 29, 2020
seisman
pushed a commit
that referenced
this pull request
Aug 1, 2020
* Try to use filelock for downloads We want to avoid race conditions on gmt_data_server.txt and gmt_hash_server.txt. * Update gmt_remote.c * Delete a failing gmt_data_server.txt file so it can be regenerated If for whatever reason the gmt_data_server.txt ends up blank, reading will fail. Rather than just complain about it, we now delete the file since we (a) know it is broken and (b) unless it is removed we will not attempt to refresh it for another cycle (by default 24 hours).
seisman
added a commit
that referenced
this pull request
Aug 1, 2020
* Let libcurl downloads be managed via lockfiles (#3735) * Try to use filelock for downloads We want to avoid race conditions on gmt_data_server.txt and gmt_hash_server.txt. * Update gmt_remote.c * Delete a failing gmt_data_server.txt file so it can be regenerated If for whatever reason the gmt_data_server.txt ends up blank, reading will fail. Rather than just complain about it, we now delete the file since we (a) know it is broken and (b) unless it is removed we will not attempt to refresh it for another cycle (by default 24 hours). * Exempt URL queries from having lock files (#3768) Since the URL of a quiery is not a good file name, we do not create an advisory lock file for such downloads. Addresses #3765 hopefully. * Checked wrong string for URL (#3780) Hopefully fixes #3765. * Forgot the other place where locking occurred (#3785) Now both places where lockfiles are used avoids URL queries. Closes #3765. * Must close file before delete it (#3804) Co-authored-by: Paul Wessel <[email protected]> Co-authored-by: Joaquim <[email protected]>
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.
Description of proposed changes
When the time is up to refresh the gmt_data_server.txt file, and you run several commands, possibly simultaneously, that all need to access remote files (not necessarily the same file), they all realize that the local copy of gmt_data_server.txt is too old and must be refreshed. So then the mad dash of being the first to do this starts among competing processes. What seems to happen is that they are clobbering each other and sometimes we are left with a zero-size gmt_data_server.txt file. This file now has a fresh date, so it is no longer downloaded, but it has no contents so any remote grids will fail to be aquired. This problems seems particulary bad when I run ctest with 20+ cores and that file is due for a refresh.
This PR tries to use the lockfile mechanism we have used for years for the gmt.history file (to protect them from being clobbered with your run process1 | process2, for instance. Except here, the file is acquired inside libcurl. The solution explored in this PR is to start the download sequence as outlined in #3723. So far so good. I did two experiments:
I hope @joa-quim and @seisman can give this branch a spin and see if they discovered any issues. It already seems to me it works better than the no-lock release we have. I added WIP so that we dont merge until we test a bit more. Comments welcome.