hg robustcheckout taking more than 30 minutes
Categories
(Developer Services :: Mercurial: robustcheckout, defect)
Tracking
(Not tracked)
People
(Reporter: aerickson, Unassigned)
References
Details
hg robustcheckout --purge --sharebase /tmp/testdir/shared --upstream https://hg.mozilla.org/mozilla-unified --revision e9df3ce8632c777f20058835e92e9c1b99af23ed https://hg.mozilla.org/try /tmp/testdir/repo is taking 30+ minutes from a host in MDC2 and my home connection.
The data is coming from Amazon/Cloudfront. The transfers are fast initially (lots of pps and bits per second), but after each part of the clone process (changesets are fine because small, but manifests, etc) has run for a bit (not sure if it's time like 30s in or 25% done) the packets seem to start trickling in (1 pps with bursts of 100, using iptraf-ng on the MDC2 host).
Bug 1672295 has additional details.
|
Reporter | |
Comment 1•4 years ago
|
||
Clones take ~15-20 minutes with mercurial 4.7.1 (4.8.1 seems ok also) and 40+ with mercurial 5.5.2.
|
|
Reporter | |
Comment 2•4 years ago
|
||
Test result summary:
4.7.1: ok
4.8.1: ok
4.8.2: ok
4.9: bad
4.9.1: bad
5.5.2: bad
|
||
Comment 3•5 months ago
|
||
If this is still an issue, can you add --profile to the hg command line to run this with a current version?
FWIW I've just tried this locally on hg 6.3.2 and it took 210s.
|
|
Reporter | |
Comment 4•4 months ago
|
||
I just tested with 6.4.5 (the latest mercurial that robustcheckout supports) and got roughly the same timings across 10 runs of each 4.8.1 and 6.4.5 (~220 seconds).
Description
•