Open olivierlambert opened 6 years ago
For UPS monitoring
@mkninc thanks for the feedback, can you check if those RPMs on CentOS 7 require external deps not already present in XCP-ng?
===========================================================================================================
Package Arch Version Repository Size
===========================================================================================================
Installing:
apcupsd x86_64 3.14.14-5.el7 epel 305 k
Installing for dependencies:
libusb x86_64 1:0.1.4-3.el7 base 19 k
===========================================================================================================
Package Arch Version Repository Size
===========================================================================================================
Installing:
nut x86_64 2.7.2-4.el7 epel 1.6 M
Installing for dependencies:
freeipmi x86_64 1.5.7-2.el7 base 2.0 M
libusb x86_64 1:0.1.4-3.el7 base 19 k
nut-client x86_64 2.7.2-4.el7 epel 206 k
I love mc, totally agree with the suggestion :)
Good catch, let's close the other issue to centralize requests here :+1:
Hi, one mistake ;) Cryptosetup is cryptsetup ;)
But I have one more package as request ... nrpe
That is nice for an external montoring
yum-utils and yum-plugin-versionlock for easier package/repository management
We had added traceroute in #32 but that was in the old repo, so we need to add it again.
I usually install cryptsetup manually from the centos base repos on XS7 and it works fine.
Also:
Troobleshooting:
- htop
- iftop
- iperf
- tcpdump
- nload
- mtr
- sysstat (iostat, mpstat, sar)
- dstat
- nethogs
Monitoring:
- netdata
- net-snmp-utils
- nagios-plugins
Hardware RAID tools:
Editors
- vim
- emacs-nox
Security
Misc
locate - useful to ...locate things tmux - screen too, but tmux better for terminal sessions
bwm-ng small and simple console-based bandwidth monitor iptraf Interactive Colorful IP LAN Monitor iotop simple top-like I/O monitor shorewall(firewall)
Hi, we are heavilly using nmon - nmon.sourceforge.net have binaries for RH/CentOS 7.2, works well.
The list is becoming overwhelming and I see no reason to reject any package from it (except if they require libs that would overwrite those of XCP-ng).
I'm considering a new approach, because I don't think maintaining a significant portion of the CentOS repos in ours will make sense in the long run. Actually two new approaches:
1) Add CentOS and EPEL repos to XCP-ng but limit them to a whitelist. This means that every additional package will have to be approved manually but without the need for us to maintain the package afterwards, leaving this to RedHat and CentOS people who already do a good job.
2) Add unrestricted CentOS and EPEL repos, using the yum-plugin-priorities plugin, with the highest priority given to XCP-ng packages. This would allow installing anything from CentOS and EPEL provided they don't try to overwrite a package from XCP-ng. I'm using this yum plugin in xcp-ng-build-env for similar reasons and it seems to be working well.
In both cases I would have to rethink the yum upgrade
process a little bit because we would need to update the CentOS and EPEL repository files before the upgrade, but I think I can make it work quite easily with just an extra step to the procedure (= update xcp-ng-release before updating the other packages).
The list is becoming overwhelming and I see no reason to reject any package from it (except if they require libs that would overwrite those of XCP-ng).
True.
I'm considering a new approach. Actually two new approaches, because I don't think maintaining a significant portion of the CentOS repos in ours will make sense in the long run.
I agree.
1. Add CentOS and EPEL repos to XCP-ng but limit them to a whitelist. This means that every additional package will have to be approved manually but without the need for us to maintain the package afterwards, leaving this to RedHat and CentOS people who already do a good job.
It will require more work for the XCP-ng Team, but all packages could be tested (and upgrades with them installed too). Many question in this case:
2. Add unrestricted CentOS and EPEL repos, using the yum-plugin-priorities plugin, with the highest priority given to XCP-ng packages. This would allow installing anything from CentOS and EPEL provided they don't try to overwrite a package from XCP-ng. I'm using this yum plugin in xcp-ng-build-env for similar reasons and it seems to be working well.
It will be more difficult to reproduce problems as everybody will be able to install any packages, but people will be really free to do what they want. In this case, packages known to be working could be put on a list in the wiki, with a warning telling that others have not been tested ?
1. Add CentOS and EPEL repos to XCP-ng but limit them to a whitelist. This means that every additional package will have to be approved manually but without the need for us to maintain the package afterwards, leaving this to RedHat and CentOS people who already do a good job.
It will require more work for the XCP-ng Team, but all packages could be tested (and upgrades with them installed too). Many question in this case:
* will the list be fully open or be limited by criteria (wich ones)?
The criteria that I can think of right now: not overwriting (or depending on packages that would overwrite) existing XCP-ng packages
* who will be responsible to update this list ? update frequency ?
Me at first, then if people join to share the work this could probably be maintained collectively. Update frequency: at least once every big release, probably more often. It mainly depends on the test effort people are willing to provide.
2. Add unrestricted CentOS and EPEL repos, using the yum-plugin-priorities plugin, with the highest priority given to XCP-ng packages. This would allow installing anything from CentOS and EPEL provided they don't try to overwrite a package from XCP-ng. I'm using this yum plugin in xcp-ng-build-env for similar reasons and it seems to be working well.
It will be more difficult to reproduce problems as everybody will be able to install any packages, but people will be really free to do what they want. In this case, packages known to be working could be put on a list in the wiki, with a warning telling that others have not been tested ?
Yes
I have a specific request as the wanted package is not a tool, but a dependency for a tool.
I'm trying to install salt-minion in order to join an xcp-server to our saltstack master, which is used for hundred of servers. I follow this guide : https://repo.saltstack.com/#rhel
But salt has a few dependencies that are not satisfied :
As far as I see, It seems that systemd is not the same version as in centos classic distribution, so I think that the better solution is to add this package in xcp-ng repositories/iso. As the source package for systemd-python is systemd, I think that it can be generated with the whole systemd-* packages ?
As stormi said me on IRC, this packages is in fact available, in builddeps for 7.5, and will be in base in 7.6. So my question is already answered. Thanks to him
Adding deltarpm
as mentioned in #120
nvme-cli
, as mentioned in https://xcp-ng.org/forum/topic/908/smartmontools
Netdata. For the love of data.
yum-cron
, automatic security updates
joe
- A simple text editor
Current wishlist, with additional information for each package.
package | provenance | added dependencies | action |
---|---|---|---|
htop | epel | none | installed by default (8.0) |
iftop | epel | none | installed by default (8.0) |
iperf | epel | none | added to repo (8.0) |
iperf3 | centos | none | added to repo (8.0) |
traceroute | centos | none | added to repo (8.0) |
already installed | none | - | |
nload | epel | none | added to repo (8.0) |
mtr | centos | none | added to repo (8.0) |
already installed | none | - | |
dstat | centos | none | added to repo (8.0) |
nethogs | epel | none | |
bwm-ng | epel | none | |
iptraf-ng | centos | none | |
iotop | centos | none | added to repo (8.0) |
nmon | epel | none | |
pv | epel | none | |
fio | centos | several |
package | provenance | added dependencies | action |
---|---|---|---|
netdata | needs packaging | ? | |
nrpe | epel | nagios-common | |
check-mk-agent | epel | GeoIP,bind-libs,bind-utils,bind-libs-lite,bind-license | |
munin-node | epel | 62 packages | |
net-snmp-utils | centos | net-snmp,net-snmp-agent-libs,net-snmp-libs | |
nagios-plugins | epel | nagios-common |
package | provenance | added dependencies | action |
---|---|---|---|
wireguard | needs packaging | ? | |
cryptsetup | centos | cryptsetup-libs | installed by default (8.0) |
fail2ban | epel | a lot | |
ferm | epel | none |
package | provenance | added dependencies | action |
---|---|---|---|
tw_cli | needs packaging | ? | |
storcli | needs packaging, a RPM exists on broadcom's website | none at runtime |
package | provenance | added dependencies | action |
---|---|---|---|
apcupsd | epel | libusb | |
nut | epel | freeipmi,libtool-ltdl,libusb,nut-client |
package | provenance | added dependencies | action |
---|---|---|---|
already installed | none | - | |
dc3dd | epel | none | |
nvme-cli | centos | none | added to repo (8.0) |
lshw | centos | none | added to repo (8.0) |
package | provenance | added dependencies | action |
---|---|---|---|
vim-enhanced | centos | gpm-libs,vim-common,vim-filesystem | added to repo (8.0) |
emacs-nox | centos | 8 packages | |
joe | epel | none | added to repo (8.0) |
package | provenance | added dependencies | action |
---|---|---|---|
deltarpm | centos | none | |
dnsutils | needs packaging | ? | |
mlocate | centos | none, but doesn't it add I/O overhead? | |
mc | centos | gpm-libs | added to repo (8.0) |
socat | centos | none | added to repo (8.0) |
tmux | centos | none | added to repo (8.0) |
wpa_supplicant | centos | none | added to repo (8.0 updates) |
yum-cron | centos | none | available in repo as part of yum suite, but I don't advise using it! (8.0) |
yum-utils | centos | none | installed by default (8.0) |
yum-plugin-versionlock | centos | none | available in repo as part of yum suite, but I don't advise using it! (8.0) |
Suggested on the forum: socat, useful to open VNC access
For netdata, there are official packages now : https://packagecloud.io/netdata/netdata
dc3dd is usefull to wipe a disk (it is like dd if=/dev/zero but with live status & more options)
For netdata, there are official packages now : https://packagecloud.io/netdata/netdata
and netdata is present in epel repo too.
Problem is that netdata official package doesn't have Xen stuff IMHO
A small comment since the last comment above isn't true anymore: we now provide netdata, either through Xen Orchestra (hosts will stream the data towards XO, that's the recommended setup because it doesn't write anything on host disks and uses little memory) or as a standalone installation (netdata-ui
from our repositories).
Link Layer Discovery Protocol packages lldpd or lldpad. My hunch is that lldpd is better and provides lldpcli, but have not used them as much.
lldpad
is available as a CentOS package and has no additional external dependency so I could include it easily.lldpd
is available as an EPEL package and has no additional external dependency so I could include it easily too.The OS performance tuning package "tuned" provides specific options. Would "virtual-host" optimize performance of xcp-ng like it does for KVM?
$ tuned-adm list Available profiles: - balanced - General non-specialized tuned profile - desktop - Optimize for the desktop use-case - hpc-compute - Optimize for HPC compute workloads - latency-performance - Optimize for deterministic performance at the cost of increased power consumption - network-latency - Optimize for deterministic performance at the cost of increased power consumption, focused on low latency network performance - network-throughput - Optimize for streaming network throughput, generally only necessary on older CPUs or 40G+ networks - powersave - Optimize for low power consumption - throughput-performance - Broadly applicable tuning that provides excellent performance across a variety of common server workloads - virtual-guest - Optimize for running inside a virtual guest - virtual-host - Optimize for running KVM guests
Suspect tuned-adm still requires hand-holding when executed on CentOS as mentioned by this RHEL employee in Boosting CentOS server performance
I'm not sure that tuned tool would be adequate for dom0. dom0 is already supposed to be optimized for virtualization. If it isn't, that is something we should fix in dom0 directly, not rely on users running an external tool to tweak things and risk break it.
Yes, tuned would be strictly for testing to find out which parameters to investigate. So would it even make sense to have as a permanent member of a testing repository and never make it to a normal repository.
Yes, tuned would be strictly for testing to find out which parameters to investigate. So would it even make sense to have as a permanent member of a testing repository and never make it to a normal repository.
Then you can consider that it's already the case since you can install it with a single command:
yum install tuned --enablerepo=base,updates
This will pull several dependencies and I clearly discourage anyone from doing it on a production server.
* `lldpad` is available as a CentOS package and has no additional external dependency so I could include it easily. * `lldpd` is available as an EPEL package and has no additional external dependency so I could include it easily too.
Actually lldpad is already present in XCP-ng and installed by default.
lldpcli in lldpd is much more useful for the client side ops.
`lldpcli neighbors`
On Wed, Apr 29, 2020 at 9:23 AM Samuel VERSCHELDE notifications@github.com wrote:
lldpad
is available as a CentOS package and has no additional external dependency so I could include it easily.
lldpd
is available as an EPEL package and has no additional external dependency so I could include it easily too.Actually lldpad is already present in XCP-ng and installed by default.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/xcp-ng/xcp/issues/56#issuecomment-621244062, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACX7FZQXUOBNLF4OGVKXQLRPAZ53ANCNFSM4FRAM2CA .
lldpad in XCP-ng is patched, with a patch named dont-add-xs-vif-to-lldpad.patch
. Which makes me wonder if lldpd
could need such patching in order not to cause any harm when used.
Hi,
please add:
apcupsd
nut
would be great so I could control my eaton ups :-)
I'm planning to do it... Sooner or later :)
OK cool thanks!
initramfs-tools and/or dracut if possible.
What's the benefit for XCP-ng? (What do you need/use it for?)
dracut is already present
Any chance of adding 'tuptime'? It's a python script for monitoring and reporting on uptime.
Any chance of adding 'tuptime'? It's a python script for monitoring and reporting on uptime.
Hi. Probably not, unless many users ask for it, because I can't find any package named tuptime
in CentOS or EPEL repositories.
Any chance of adding 'tuptime'? It's a python script for monitoring and reporting on uptime.
Hi. Probably not, unless many users ask for it, because I can't find any package named
tuptime
in CentOS or EPEL repositories.
There are some packages here: https://copr.fedorainfracloud.org/coprs/frankcrawford/tuptime/
Not sure about their status.
Andy
We should be able to allow people to download directly some "extra" packages that can be really useful.
The only downside is that we'll need to synchronize any package update from CentOS to XCP-ng.
Monitoring
General
top
): no extra depsSecurity
UPS
Misc
If you have ideas of stuff that you'll like, let us know here