Jason Zhao wrote:
Hi, Shawn,
Jason Zhao wrote:
Hi, Experts,
I am doing our local pkg server, and the pkg.depotd(1m) advise
me to run "--refresh-index", and it seems like it tries
to update index directory on IPS repo. But when I run the
command:
# pkg.depotd -d <my/repo> --refresh-index.
I found the process will exit immediately after I run it, but
it tries to fork millions of times of "--refresh-index". All
the commands(new processes) exit immediately.
My question are:
- What's usage of "--refresh-index", I could not find it in manual.
It's not in the manual because a user using that at the command line
isn't supported. It's a flag that tells the depot it's been started by
another depo for the express purpose of rebuilding the index, then dying.
- Why does the above command fork so much times on itself? Maybe one
running time would be enough. If the command fails, it should
print detailed information other than repeatedly fork itself.
Despite that, this shouldn't happen. If you can reproduce it, please let
me know as I'd like to get this fixed asap.
[snip
One more question,
Right now, the "pkg.depotd -d <my/ips/repo> --refresh-index could
create a directory "TMP"
on "index" directory, and create some files as following.
# cd /my/repo/index/TMP
# ls
__at_depend __at_set
__st_legacy_pkg sort.0 sort.7
__at_directory __st_None
__st_license sort.1 sort.8
__at_driver __st_alias
__st_path sort.10 sort.9
__at_file __st_basename
__st_srcpkgs sort.2
token_byte_offset.v1
__at_hardlink __st_content
__st_variant.arch sort.3
__at_legacy __st_description
__st_variant.zone sort.4
__at_license __st_driver_name
main_dict.ascii.v2 sort.5
__at_link __st_info.classification
pkg sort.6
Then I tried to restart IPS server manual, it shows errors.
# usr/lib/pkg.depotd -d /my/repo -p 10000
[26/Mar/2009:14:25:09] ENGINE Listening for SIGHUP.
[26/Mar/2009:14:25:09] ENGINE Listening for SIGTERM.
[26/Mar/2009:14:25:09] ENGINE Listening for SIGUSR1.
[26/Mar/2009:14:25:09] ENGINE Bus STARTING
[26/Mar/2009:14:25:09] ENGINE Started monitor thread '_TimeoutMonitor'.
[26/Mar/2009:14:25:09] ENGINE Serving on 0.0.0.0:10000
[26/Mar/2009:14:25:09] ENGINE Bus STARTED
[26/Mar/2009:14:25:12] INDEX Updating search indices
[26/Mar/2009:14:25:14] INDEX Index corrupted, remove all files and
rebuild from scratch by clearing out /ips/repo/repo/index and
restarting the depot.
[26/Mar/2009:14:25:14] INDEX ERROR building search database, exit
code: 1
Right now, the IPS server could start, but tt hints the "Index
corrupted", is that an issue or obstacle for "pkg
install/image-update"?
If yes, how can I fix it?
Are usr/lib/pkg.depotd and pkg.depotd pointing to the same binaries/pkg
versions? If that, that could easily be your problem. Search index
issues shouldn't prevent you from being able to use the depot to install
and or image update. In other words, if the indexes are broken, it won't
server search, but everything else should work as expected.
[snip]
Right now, our local IPS repo has been started, and we are making
testing on 2009.06 AI release. And we could not stop the IPS repo
now. But I will try if I can reproduce it after we finish the test cycle.
That's maybe several days later.
Please let me know if you're able to reproduce this behavior. I've seen
it before, but put back a fix that prevented it in the instance that I'm
aware of.
Thanks
Jason
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
Thanks,
Brock
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
