Skip to content

first step to keep task attributes for staging tasks and time window support for async log grep - #850

Merged
tmaeno merged 4 commits into
PanDAWMS:masterfrom
tmaeno:master
Nov 3, 2026
Merged

tmaeno merged 4 commits into
PanDAWMS:masterfrom
tmaeno:master

Conversation

@tmaeno

@tmaeno tmaeno commented Nov 3, 2026

Copy link
Copy Markdown
Contributor

No description provided.

tmaeno and others added 4 commits November 2, 2026 21:23
A grep reaches the recent part of a live log only with tail_bytes, and
cold disks read 7 to 33 MB/s, so a tail of 3 GiB often outlasts the
240 s a grep may run and returns nothing: what a busy panda-DBProxy.log
took more than about three hours ago could not be read before the next
morning's .gz.

submit_grep_request takes since and until, in UTC as the log stamps are,
for a plain file with whole_records. The processor bisects the file on
the time stamps of its records to find the bytes that hold the window,
60 s wider on each side for records written out of order, and passes
only those to rg, from a thread that reads the file it bisected. The
window in the result says where the bytes searched start and end. One of
more than 1 GiB is cut, which window.cut says rather than truncated, as
it calls for asking further rather than differently. tail_bytes goes
through the same reader instead of tail.
Commented out the conditional check for 'staged' task status.
@tmaeno
tmaeno merged commit 70400a8 into PanDAWMS:master Nov 3, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant