Description of problem:
According to bug 1368420, due to rhv-h update rpm is large, it maybe failed to upgrade RHVH via RHN if the network is not good, so suggest rhv-h to reduce deptree to reduce update rpm.
For example, we can get rid of unnecessary packages, such as x11 etc.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
(In reply to Huijuan Zhao from comment #0)
> Description of problem:
> According to bug 1368420, due to rhv-h update rpm is large, it maybe failed
> to upgrade RHVH via RHN if the network is not good, so suggest rhv-h to
> reduce deptree to reduce update rpm.
> For example, we can get rid of unnecessary packages, such as x11 etc.
"rhv-h update rpm is large" should be "rhv-h image is large", so it maybe failed
to upgrade rhv-h via CDN if the network speed is not good.
I think we have already a bug for that (or it is the appliance image?). Certainly if we can remove groups (such as X11? why is it indeed there?) it'd be great, but I would not spend more than that - this is exactly blacklisting we worked too hard on in previous RHEVH.
We examined using deltarpms for both appliance and RHV-H, but deltarpm doesn't seem to understand how to compare disk images, and the delta is actually larger than the initial RPM. This may be a good target for an RFE.
We're not actually able to reduce the size of RHV-H in this way, though. Our list of dependencies is extremely small (see below). Any other package is pulled in as a dependency of these somewhere.
libX11/gtk2, for example, are pulled in by redhat-release-virtualization-host-content->vdsm->virt-v2v->(pile of X11 libs)
In order to effectively reduce the size, we need to file RFEs against the appropriate components to get the deps dropped.
The only low-hanging fruit is dropping wireless firmwares, which saves ~25MB, or 5% of the size, and is something we can/should do from the RHV-H side.
%packages --excludedocs --ignoremissing
# The RHV-H Entry Point Package
# Anaconda in RHEL assumes that openscap is available,
# thus we need to include it as well.
# Bug-Url: https://bugzilla.redhat.com/show_bug.cgi?id=1331354
# Core RHV requirements
# Additional RHV requirements
# Enable FIPS
# Requirements to enable services
For 4.1.2 RC2:
ovirt-node-ng-image-4.1.2-0.2.rc2.el7.centos.noarch.rpm 529.29 MB
Latest master / 4.2 build:
ovirt-node-ng-image.squashfs.img 535.96 MB
ovirt-node-ng-image-4.2.0-0.3.master.20170420000146.git7f66ae2.el7.centos.noarch.rpm 536.05 MB
So we're moving in the wrong direction.
Can you please check?
The following packages were added in master (total of 20M installed)
abrt-2.1.11-45.el7.centos (size: 2432772)
abrt-addon-ccpp-2.1.11-45.el7.centos (size: 356023)
abrt-addon-kerneloops-2.1.11-45.el7.centos (size: 38349)
abrt-addon-python-2.1.11-45.el7.centos (size: 31462)
abrt-addon-vmcore-2.1.11-45.el7.centos (size: 42730)
abrt-dbus-2.1.11-45.el7.centos (size: 132116)
abrt-libs-2.1.11-45.el7.centos (size: 57696)
abrt-python-2.1.11-45.el7.centos (size: 57771)
abrt-retrace-client-2.1.11-45.el7.centos (size: 104659)
elfutils-0.166-2.el7 (size: 775048)
gdb-7.6.1-94.el7 (size: 7336957)
libreport-2.1.11-35.el7.centos (size: 2116187)
libreport-plugin-rhtsupport-2.1.11-35.el7.centos (size: 106273)
libreport-plugin-ureport-2.1.11-35.el7.centos (size: 57134)
libreport-python-2.1.11-35.el7.centos (size: 96467)
libreport-web-2.1.11-35.el7.centos (size: 45296)
libtar-1.2.11-29.el7 (size: 74934)
lshw-B.02.17-12.el7 (size: 858143)
pygobject2-2.28.6-11.el7 (size: 890111)
python2-gluster-3.10.1-1.el7 (size: 16711)
python2-passlib-1.6.5-1.el7 (size: 2882209)
python-augeas-0.5.0-2.el7 (size: 72850)
python-prettytable-0.7.2-2.el7.centos (size: 183858)
satyr-0.13-14.el7 (size: 2396229)
xmlrpc-c-1.32.5-1905.svn2451.el7 (size: 330001)
xmlrpc-c-client-1.32.5-1905.svn2451.el7 (size: 53400)
abrt and lshw were added as direct dependencies for vdsm.
The abrt bunch pulled in gdb, elfutils, satyr, python-augeas, pygobject2 and libreport which pulled xmlrpc-c and libtar as well.
python-passlib was pulled in by ansible.
python2-gluster and python-prettytable were pulled in by glusterfs-geo-replication which is required by vdsm as well.
Closing can't fix.
We got new packages related to metrics and katello agent to be added in 4.2, so we can only increase the size, not reduce it.