[LU-3492] Lustre 2.4 SRPM Re-Build Fails Created: 21/Jun/13 Updated: 09/Oct/21 Resolved: 09/Oct/21 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.4.0 |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Major |
| Reporter: | Malcolm Cowe (Inactive) | Assignee: | WC Triage |
| Resolution: | Low Priority | Votes: | 0 |
| Labels: | llnl | ||
| Issue Links: |
|
||||||||
| Severity: | 3 | ||||||||
| Rank (Obsolete): | 8778 | ||||||||
| Description |
|
When attempting to build Lustre Client RPM packages for Lustre 2.4, the build fails with the following error: ... Making all in utils RPM build errors: The commands used are as follows: wget -r -np --level=0 http://downloads.whamcloud.com/public/lustre/lustre-2.4.0/el6/client/SRPMS/ cd downloads.whamcloud.com/public/lustre/lustre-2.4.0/el6/client/SRPMS/ rpmbuild --rebuild --define 'lustre_name lustre-client' \ System is CentOS 6.4, running update kernel 2.6.32-358.6.2.el6.x86_64. Any ideas? I've used an identically similar process for previous releases of Lustre without issue. I get a similar error when building from the v2_4_0 tag in Git, using ./configure ... && make. |
| Comments |
| Comment by Malcolm Cowe (Inactive) [ 24/Jun/13 ] |
|
It appears that by explicitly requesting the server build to be disabled, that the SRPM rebuild can be completed. Similarly for running a build on master or the v2_4_0 tag on the Git source. The command line I used to verify is: rpmbuild --rebuild --define 'lustre_name lustre-client' --define 'configure_args --disable-server' lustre-client-2.4.0-2.6.32_358.6.2.el6.x86_64_gd3f91c4.src.rpm and also: ./configure --enable-client --disable-server for a build of the Git source. This is a change in behaviour compared with previous releases of Lustre. |
| Comment by Jodi Levi (Inactive) [ 24/Jun/13 ] |
|
Brian, |
| Comment by Brian Murrell (Inactive) [ 24/Jun/13 ] |
|
Not much to say without being able to dig into it, but it looks like trying to build the client is for some reason trying to build ldiskfs, which it shouldn't be. There have been other reports of needing to explicitly set --disable-server which never used to be the case. At one time, configure figured out if the build was for server or client by itself, so some regression with that autodetection has been introduced. |
| Comment by Christopher Morrone [ 25/Jun/13 ] |
|
The problem is that Intel is building a src rpm named lustre-client-<verion>.src.rpm, which is in reality almost exactly the same thing as lustre-<version>.src.rpm. But obviously, users of the rpms are going to think that the one with the name "lustre-client" is only the lustre client. This is bad by design, and needs to go away. There should be just one src rpm for each spec file, that can be rebuilt reasonably using good, clean options to rpmbuild --rebuild. In other words, the "lustre-client" source rpm is not an rpm of the lustre client code. It is an rpm of all of lustre. So Malcolm, what you are doing right now is exactly correct, and expected. But I too agree that this is bad, and I believe that this problem would be addressed by |