Bug 440190
| Summary: | PackageKit deadlocks with two concurrent installs | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Jeremy Fitzhardinge <jeremy> | ||||||||||
| Component: | PackageKit | Assignee: | Robin Norwood <robin.norwood> | ||||||||||
| Status: | CLOSED RAWHIDE | QA Contact: | Fedora Extras Quality Assurance <extras-qa> | ||||||||||
| Severity: | medium | Docs Contact: | |||||||||||
| Priority: | low | ||||||||||||
| Version: | rawhide | CC: | lordmorgul, richard, tim.lauridsen | ||||||||||
| Target Milestone: | --- | ||||||||||||
| Target Release: | --- | ||||||||||||
| Hardware: | All | ||||||||||||
| OS: | Linux | ||||||||||||
| Whiteboard: | |||||||||||||
| Fixed In Version: | Doc Type: | Bug Fix | |||||||||||
| Doc Text: | Story Points: | --- | |||||||||||
| Clone Of: | Environment: | ||||||||||||
| Last Closed: | 2008-04-15 16:53:30 UTC | Type: | --- | ||||||||||
| Regression: | --- | Mount Type: | --- | ||||||||||
| Documentation: | --- | CRM: | |||||||||||
| Verified Versions: | Category: | --- | |||||||||||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||||||||
| Cloudforms Team: | --- | Target Upstream Version: | |||||||||||
| Embargoed: | |||||||||||||
| Bug Depends On: | |||||||||||||
| Bug Blocks: | 441857 | ||||||||||||
| Attachments: |
|
||||||||||||
|
Description
Jeremy Fitzhardinge
2008-04-02 06:13:30 UTC
Do the two installs complete eventually? No, at least not before I rebooted about 30 mins later. The BackupPC RPM is in /var/cache/yum/packages/rawhide, but there's no sign of cabextract. I'm doing an update at the moment (~60 packages). The pk applet is just showing "updating package audit-libs (downloading)" - and has been for some time. However it appears it's actually downloading all the packages, since the network is busy and they're appearing in /var/cache/yum/rawhide/packages. I wonder if BackupPC had a number of dependent packages which it was downloading, but I was confused because there was no indication that was happening. In other words, perhaps this bug should be something along the lines of "packagekit applet should show incremental progress during updates/installs". I'm seeing something similar with PK trying to get update information 4 times at once. Just before this I had an install fail to complete like above comments, then rebooted, checked my rpmdb which was corrupt (screen to come) and had to do: rm /var/lib/rpm/__db*; rpm --rebuilddb Afterward, I'm seeing these multiple operations occur at once. I did an update manually via Update System, but as soon as it completed I clicked the Refresh Info button and it never completed. The 4 copies of refresh were found then as the UI was stuck. Created attachment 300183 [details]
pk-update-exception.png
Update error, exception during manually requested update operation.
Created attachment 300184 [details]
rpmdb broken after install failure
Created attachment 300185 [details]
PK-multiple-refresh-ops-2.png
Showing refresh operations attempting to happen at once, I tried canceling
these by opening each and selecting cancel one at a time, but none of them
managed to stop. The VM eventually locked up completely.
I'm seeing this issue continue on both i686 and x64 where about once a day I get operations hanging (both yum and packagekit) due to the corrupted __db.xxx files. Removing them temporarily fixes the problem but it keeps returning. I cloned this as a separate bug since its a different issue than original report and I should have started a new one in the first place. PK does do some updates fairly slowly but I think its a mirror selection problem for me when that happens (no fastestmirror plugin yet). Sorry for the spam with the two bugs though. Created attachment 302135 [details]
in git
Jeremy, we've added UI like this to show you when we have to add new deps to a
transaction to allow it to download. Would this solve your bug?
It would certainly give more information to distinguish between an unexpectedly large install (due to dependencies), and an actual hang. Marking as fixed, as the status of this bug is unclear, and the original big has been fixed. Please open a specific bug if you think there is still a problem. Thanks. |