-
Notifications
You must be signed in to change notification settings - Fork 61
Comparing changes
Open a pull request
base repository: cea-hpc/robinhood
base: master
head repository: stanford-rc/robinhood
compare: main
- 15 commits
- 19 files changed
- 2 contributors
Commits on Jul 29, 2026
-
param_utils.c: add support for creation_time placeholder
Add creation_time as valid placeholder in action_params values. Change-Id: I5fb0a899fc6bb23e82bbff4c5affb326194a72e5 Signed-off-by: Stephane Thiell <sthiell@stanford.edu>
Configuration menu - View commit details
-
Copy full SHA for 1d12ae4 - Browse repository at this point
Copy the full SHA 1d12ae4View commit details -
policies: add optional stripe_index prefix for ost_index
With this patch, ost_index can now take a an optional "stripe_index:" prefix to match ost_index on the specified OST stripe index only, instead of checking for ost_index on any OST stripe. This allows for a policy to run on multiple Robinhood instances based on Lustre the OST striping distribution without risk of collision (see LU-17170). For example, with this option enabled, when using the following rule: rule default { condition { (creation > 2d and ((ost_index == 0:96) or (ost_index == 0:97) or (ost_index == 0:98) or ...) Only files what have their first stripe on OSTs 96, 97 and 98 will be considered. Other instances can filter different first OSTs on the same stripe index to make sure they don't overlap. Change-Id: I402340dcc160ace876d64c3d030153edd079088aConfiguration menu - View commit details
-
Copy full SHA for 695386d - Browse repository at this point
Copy the full SHA 695386dView commit details -
Change-Id: I43ca0d72d23da5920d6b5076dee1d650b4a797e4
Configuration menu - View commit details
-
Copy full SHA for 3303ce4 - Browse repository at this point
Copy the full SHA 3303ce4View commit details -
Change-Id: Ib30063d1d692d7d948cccd7737389a41b8ce7e14
Configuration menu - View commit details
-
Copy full SHA for 5f539c9 - Browse repository at this point
Copy the full SHA 5f539c9View commit details -
Change-Id: I06dbabb8c1f8748a5dc57403cc38d1e320e201bf
Configuration menu - View commit details
-
Copy full SHA for b1c6f85 - Browse repository at this point
Copy the full SHA b1c6f85View commit details -
Change-Id: Ia0273cce40958458f97f6b73d66c988ba9edb038
Configuration menu - View commit details
-
Copy full SHA for 3873af1 - Browse repository at this point
Copy the full SHA 3873af1View commit details -
spec: reinstate tests with only doc
Change-Id: Ie4fdf1fb0592f54461d0b4af5a1e70ac5873f0a1
Configuration menu - View commit details
-
Copy full SHA for 5a20808 - Browse repository at this point
Copy the full SHA 5a20808View commit details -
Change-Id: Ic7dc895e8a8cc918f185913afb8d5448350b3ddf
Configuration menu - View commit details
-
Copy full SHA for 6e3b315 - Browse repository at this point
Copy the full SHA 6e3b315View commit details -
cfg: parse ost_index values as strings (stripe_index prefix)
Complete the commit "policies: add optional stripe_index prefix for ost_index": criteria2filter() and eval_condition() were changed to read ost_index values as strings, but the config parser still validated them as positive integers, rejecting the "stripe_index:ost_index" form with 'integer expected'. Store ost_index criteria as PT_STRING, print it as a string in printable conditions, and drop the early-break in eval_condition()'s OST loop. Recovered verbatim from the robinhood-debugsource-3.1.7.6-1 RPM: these hunks were present in the production build tree but never committed.
Configuration menu - View commit details
-
Copy full SHA for c59443c - Browse repository at this point
Copy the full SHA c59443cView commit details -
common: spawn policy actions with posix_spawn when available
g_spawn_async_with_pipes() forks: the atfork handlers take every malloc arena lock and the kernel copies the parent's page tables, stalling all worker threads for the duration of each fork and capping action throughput per daemon regardless of nb_threads. When posix_spawn_file_actions_addclosefrom_np() is available (glibc >= 2.34, detected at configure time), spawn children with posix_spawnp() instead: no atfork handlers, no page-table copy, same semantics (PATH search, piped stdout/stderr, GLib child watch, no fd inheritance). Older glibc (e.g. el7) keeps the g_spawn path, where posix_spawn would fall back to fork() internally anyway.
Configuration menu - View commit details
-
Copy full SHA for b88860c - Browse repository at this point
Copy the full SHA b88860cView commit details -
Configuration menu - View commit details
-
Copy full SHA for 321f869 - Browse repository at this point
Copy the full SHA 321f869View commit details -
Adopt the upstreamversion-srccX release convention.
Configuration menu - View commit details
-
Copy full SHA for 0c9c2f4 - Browse repository at this point
Copy the full SHA 0c9c2f4View commit details -
policy: use raw iterator count for end-of-list
ListMgr_GetNext() can skip rows that disappear between the candidate query and the attribute lookup. Use the number of rows selected by the iterator, rather than only the number delivered to the policy loop, when deciding whether another batch is needed. Fixes #154.
Configuration menu - View commit details
-
Copy full SHA for 229e859 - Browse repository at this point
Copy the full SHA 229e859View commit details -
policy: add primary-key cursor for unsorted batches
Add an opt-in ENTRIES.id cursor for split, unsorted policy queries. Continue each full batch with an exclusive primary-key boundary, and reject unsupported iterator shapes and deleted-entry policies.
Configuration menu - View commit details
-
Copy full SHA for 714604a - Browse repository at this point
Copy the full SHA 714604aView commit details -
Feeder end-of-list fix (#154) and opt-in primary-key cursor for split candidate queries.
Configuration menu - View commit details
-
Copy full SHA for 9d43571 - Browse repository at this point
Copy the full SHA 9d43571View commit details
This comparison is taking too long to generate.
Unfortunately it looks like we can’t render this comparison for you right now. It might be too big, or there might be something weird with your repository.
You can try running this command locally to see the comparison on your machine:
git diff master...main