This is not necessarily the right thing.
For example, nanoHUB submits batches of jobs that are mixtures
of speculative and user-submitted;
these simply need to be given low and high priorities.
The way things work now:
- if you want allocation-based prioritization,
set the "allocation_priority" flag in the request object
(Python or PHP API)
- or set the "priority" field in the request object;
that sets the priority of all the jobs in the batch
- or set the "priority" field of jobs in the batch.
Updated wiki docs accordingly.
Also fixed a bug in the test script
(note to self: adding an object to an array adds a reference, not a copy).
I have no idea why, but out of nowhere ReCaptcha stopped working on both SU and S@h.
Changing an arg to the recaptcha constructor
(from a Stack Overflow entry from 2015) fixed the problem.
It's useful to see the stats (mean, max) of memory and disk usage,
e.g. so that you can set the appropriate bounds for future jobs.
In the case of universal VM apps (e.g. nanoHUB@home)
the natural place to put this is at the batch level;
jobs in a a batch use the same tool.
Also: add a general function for printing sizes in MB and GB,
and move a function from util.inc to the appropriate file.
* Less technical error messages, reduced code duplication, changed makeproject to generate keys, added op php file for generating keypairs, changed the button styling, changed input to textarea.
* Split the ops scripts into separate check and generate files, reduced code duplication by creating a new inc file to define key variables.
Scenario:
- a submit_batch() creates a batch in INIT state
and starts creating jobs
- while it's in progress, a query_batches() RPC enumerates batches
- while the query_batches() is in progress,
the submit_batch() finishes and changes the state to IN_PROGRESS
- the query_batches() finishes and rewrites the batch in INIT state
Solution: query_batches() doesn't update batches in INIT state
Old: the remote job submission log had a hardwired name,
it was being put in possibly the wrong directory,
and disabling logging required editing a PHP file.
New: logging is enabled by putting
<remote_submit_log>filename</remote_submit_log>
in your config.xml; the file is put in the <log_dir> directory.
Also: you can log the XML request messages
(useful mainly for debugging) by putting
<remote_submit_request_log>filename</remote_submit_request_log>
in config.xml
Documented both of these in a new file:
https://boinc.berkeley.edu/trac/wiki/RemoteLogs
* Added a new config.xml setting "account_creation_rpc_require_consent" (default: disabled)
* Added condition to create_account RPC to return an error if consent details aren't given but required
* Defined a dedicated error number and default message (for C and Android use)
David Anderson