Bug 1269086
Summary: | regression in pxe install. very slow pxe installation (stall for an hour) | ||
---|---|---|---|
Product: | Red Hat Enterprise Linux 7 | Reporter: | Levente Farkas <lfarkas> |
Component: | python-urlgrabber | Assignee: | Valentina Mukhamedzhanova <vmukhame> |
Status: | CLOSED DUPLICATE | QA Contact: | BaseOS QE - Apps <qe-baseos-apps> |
Severity: | unspecified | Docs Contact: | |
Priority: | unspecified | ||
Version: | 7.1 | CC: | lfarkas, mkolman, packaging-team-maint, pmancillas |
Target Milestone: | rc | ||
Target Release: | --- | ||
Hardware: | Unspecified | ||
OS: | Unspecified | ||
Whiteboard: | |||
Fixed In Version: | Doc Type: | Bug Fix | |
Doc Text: | Story Points: | --- | |
Clone Of: | Environment: | ||
Last Closed: | 2016-01-14 12:08:20 UTC | Type: | Bug |
Regression: | --- | Mount Type: | --- |
Documentation: | --- | CRM: | |
Verified Versions: | Category: | --- | |
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
Cloudforms Team: | --- | Target Upstream Version: | |
Embargoed: | |||
Attachments: |
Description
Levente Farkas
2015-10-06 09:34:00 UTC
Please attach the logs from /tmp/*log asa individual text/plain attachments. Also, when it gets stuck switch to the shell and attach the output from: lsof -p <pid of anaconda-yum> strace -o anaconda-yum.strace -p <pid of anaconda-yum> You'll have to crtl-c the strace, let it gather for a minute or so. Created attachment 1080525 [details]
anaconda lsof during wait
Created attachment 1080526 [details]
anaconda yum strace during wait
Created attachment 1080527 [details]
after install log file in tmp anaconda
Created attachment 1080528 [details]
after install log file in tmp ifcfg
Created attachment 1080529 [details]
after install log file in tmp lsof
Created attachment 1080530 [details]
after install log file in tmp packaging
Created attachment 1080531 [details]
after install log file in tmp program
Created attachment 1080532 [details]
after install log file in tmp storage
Created attachment 1080533 [details]
after install log file in tmp strace
Created attachment 1080534 [details]
after install log file in tmp X
Looking at the packaging log the single longest operation seems to be "populate transaction set" which takes about 20 minutes: 05:11:41,337 DEBUG packaging: populate transaction set 05:33:33,774 DEBUG packaging: check transaction set 05:33:33,906 DEBUG packaging: order transaction set 05:33:33,922 INFO packaging: running transaction That would certainly be something to investigate - if it really needs to take so long or if it is taking too long due to some bug. In case it really needs to take so long there should definitely be some feedback provided to the user ("populating transaction set 9%", etc.), After that you can see yum installing individual packages for about 40 minutes. Considering it is installing 1200 packages, that amounts to 30 packages being installed per minute or one every 2 seconds. This is probably I/O bound - either by the disk or network when the packages are being downloaded (IIRC the packages are continually downloaded before being installed, not at once before the transaction). one more important note is that it's working in OS-6. everything is on a gigabit local network. the tftp server, the ftp server and workstation which try to install is 8, 16 and 4gb ram and at least intel i5 cpu. so imho the local setup can't be the bottleneck. and imho if it's working (on OS-6) with yum than it should have to work on OS-7 too. yesterday i try to install my new workstation with 16gb ram 250gb ssd and gigabit network has the same effect. so imho something was totally different in OS-6 yum package management then on OS-7. i'd be happy to test or debug any kind of tipp patch enhancement. is there any progress with this regression? We had the same problem using ftp. When we changed the protocol to http the installation completed much faster. So it's definitely a problem with the ftp transfer. unfortunately it's not working for us. if we replace ftp to http it's still VERY slow. which means from a CD it takes a few second from the network it takes a half an hour (both ftp and http). sorry. it was our fault. since our http server redirect to ftp. and yes it seems only ftp protocol slow while with http it's fast again. so we've got a workaround now at least. anyway we use vsftpd for ftp. Reassigning to urlgrabber - we don't implement the fetching of anything, just give it the URLs and tell it to go. This appears to only be happening with FTP, not with HTTP. This looks like a duplicate of bug 1152628 and bug 1218272, and should be fixed in the latest version of curl. *** This bug has been marked as a duplicate of bug 1218272 *** |