Rice University logo
 
Top blue bar image
A place to post stuff
 

Posts Tagged ‘sysadmin’


Abusing Yum to Downgrade to a Specific Update in RHEL

January 28th, 2013 by chwilk

Sometimes it just doesn’t pay to keep your machines up to date…

In the case that you must force a machine back to a previous update of RHEL (e.g. latest X.Y introduces a bug, and you want to revert to X.(Y-1) ), it isn’t easy to drop an update.

In Fedora, (and copied to RHEL, but somewhat useless by default) you can use something to the effect of

yum --releasever=X-1 distro-sync

Unfortunately, this only works for major version shifts.

If you want to do this in RHEL to hit a specific point release, you must play with some repos.

These instructions assume you have a satellite and can create kickstart profiles therein.

Edit a kickstart profile and set it back to the release you want. Also select any extra repositories you may have access to (e.g. supplementary, optional) then look at the kickstart script it generates. Find the URL for install and any repo you’ve added, then create a .repo file in /etc/yum.repos.d/ for those repositories. Name each repo in a globbable format, e.g. rhel-downgrade-X-Y rhel-downgrade-X-Y-optional…

Now the moment of truth.

yum --disablerepo \* --enablerepo rhel-downgrade-X-Y\* distro-sync

You can also specify specific yum groups and individual rpms on the command line to just downgrade a particular subsystem, as long as it doesn’t have firm dependencies into the rest of the OS.

Python RHN Fun: Cleaning the Satellite

October 16th, 2012 by chwilk

Here’s a little script to delete any systems from your database that haven’t checked in for more than 90 days.

#!/usr/bin/python
# Removes systems not checking in within the last 90 days

import xmlrpclib
import getpass
import datetime

SATELLITE_URL = "http://YOUR-SATELLITE-HOSTNAME/rpc/api"
SATELLITE_LOGIN = "userid"
SATELLITE_PASSWORD = "password"

TARGET_ENTITLEMENT = ['monitoring_entitled']

if __name__ == "__main__":
    client = xmlrpclib.Server(SATELLITE_URL, verbose=0)
    SATELLITE_LOGIN = str(raw_input("RHN Userid: ")).rstrip()
    SATELLITE_PASSWORD = getpass.getpass()

    key = client.auth.login(SATELLITE_LOGIN, SATELLITE_PASSWORD)
    result_array = client.system.listInactiveSystems(key,90)
    id_array = [profile['id'] for profile in result_array]
    client.system.deleteSystems(key, id_array)
    client.auth.logout(key)

Scripting RHN Re-activations

August 3rd, 2012 by chwilk

Here’s the situation: You’ve fixed something in your RHN Satellite activation key that should have been in there, or you’ve added functionality, but when you apply the change, you’d also like to apply those changes to machines that have already been registered with that activation key. Unfortunately, there’s no nice button to make that happen in the current version.

Hopefully you’ve been using system groups to keep track of machines deployed with your activation keys, because you’ll need to load that group into the system set manager and run a small python script one each machine as root. The idea is to re-register the machines to rhn, using each machine’s custom re-activation key to ensure that they reattach to their original system profile when you do so. Note: This will work if you’ve added something to the activation key, say a new config or software channel, but it won’t take away channels you want to remove.

Here’s the script:

#!/usr/bin/python
try:
    import xmlrpclib
    import sys
    import os
    import string

    system_id = "/etc/sysconfig/rhn/systemid"

    keys = sys.argv[1:]

    if os.path.exists(system_id):
        client =  xmlrpclib.Server("http://rhn.rice.edu/rpc/api")
        keys.append(client.system.obtain_reactivation_key(open(system_id).read()))
except Exception as inst:
    # xml rpc due to  a old/bad system id
    print type(inst)
    print inst.args
    print inst
    print "Could not get reactivation key, exiting"
    sys.exit(1)
else:
    command = "rhnreg_ks --force --activationkey=" + string.join(keys, ",")
    os.system(command)
    sys.exit(0)

Python Yum Fun 2: Kickstart List

May 29th, 2012 by chwilk

Need to create a Red Hat kickstart config to duplicate the installed packages off an existing system? Want to do this without having to create a giant file listing each RPM by hand?

Here’s a python script that will pull your installed packages list and prune it down by dependencies and yum groups, and give you a (hopefully) concise list of yum groups and RPMs that will work in a kickstart config file. The only complication is that you’ll need to ensure all your yum repositories are available at kickstart time, else you may run into some RPMs that are not available.

#!/usr/bin/env python
import yum

yb = yum.YumBase()
yb.doConfigSetup()
yb.doTsSetup()
yb.doRpmDBSetup()
    
# Grab Installed Package List (ipl)
ipl=set([pkg[0] for pkg in yb.rpmdb.pkglist])
grppkgs = set([])
# Grab Group List (gl)
gl=yb.doGroupLists()
# First index of gl is Installed Group List (igl)
igl=gl[0]

# Make sure we still are accounting for Core (sometimes packages get removed)
gotcore = False
for grp in igl:
    if grp.groupid == 'core': gotcore = True

if not gotcore:
    for grp in gl[1]:
        if grp.groupid == 'core':
           igl.append(grp)

# Find leaves of dependency tree
loners = ipl.copy()
for pkg in list(loners):
    deps= yb.findDeps(yb.returnInstalledPackagesByDep(pkg)).values()[0].keys()
    for dep in [d[0] for d in deps] :
        if dep in loners:
            loners.remove(pkg)
            break

# Print a kickstartable group list first
for grp in igl:
    for pkg in grp.mandatory_packages:
        if pkg in loners:
            loners.remove(pkg)
    for pkg in grp.default_packages:
        if pkg in loners:
            loners.remove(pkg)
    print '@'+grp.groupid

# Print rpms not pruned already
for pkg in loners:
   print pkg

Python Yum Fun 1: Kickstartable Grouplist

May 17th, 2012 by chwilk

Here’s a short python code to load up the local yum database, grab the list of yum groups, and print installed and available groups by short names (which can be used in kickstart scripts more easily than the output of yum grouplist.)

#!/usr/bin/env python
import yum
import sys

if __name__ == "__main__":
    yb = yum.YumBase()
    yb.doConfigSetup()
    yb.doTsSetup()
    yb.doRpmDBSetup()
    
    # Grab Group List (gl)
    gl=yb.doGroupLists()
    # First index of gl is Installed Group List (igl)
    igl=gl[0]
    # Second index of gl is Available Group List (agl)
    agl=gl[1]
    
    # Print a kickstartable group list
    print "Installed groups:\n"
    for grp in igl:
        print "@" + grp.groupid + ": '" + grp.name + "'"

    print "Available groups:\n"
    for grp in agl:
        print "@" + grp.groupid + ": '" + grp.name + "'"


Thanks to the for publishing a great starting point to yum python programming.

PowerUp: Flattening the Learning Curve for HPC with Linux on POWER

November 29th, 2011 by chwilk

PowerUp: Flattening the Learning Curve for HPC with Linux on POWER.

ICEBOX Serial Console Adapter Pin Out

June 15th, 2010 by chwilk

Although our data center requires remote management capability for all servers, occasionally something lands that doesn’t have the ability for whatever reason. Rather than driving down to the data center every time we (and by we, I mean the researcher and I) want to test a new kernel, we decided that we needed a remote management solution.

Fortunately, we happened to have some old Linux Networkx ICE BOX node management devices (power control, temperature monitoring, and serial console over LAN) languishing in storage after the cluster they managed was retired. Unfortunately, Linux Networkx appears to be out of business, there is no Google-able documentation for the ICE BOX on line, and the serial console adapters were all thrown out when the cluster was retired.

However, thanks to an RS-232 break out box picked up from EPO, and a few hours of playing with an RJ-45 to DB9 adapter, I have a working pin-out for an ICEBOX serial terminal, and will record it here for future poor schmucks who find themselves in my position.

I only worked out the TD, RD, and SG lines, if you need flow control (and know how to make the ICEBOX use it!) you’re on your own.

From the RJ-45 end, I used wires 2, 3, and 5 (according to the accounting on, and using the same RJ-45 to DB9 adapter as Dan Gottesman’s website, http://www.ossmann.com/5-in-1.html).

RJ-45 Signal DB-9 Signal
2 (orange) RD 3 TD
3 (black) TD 2 RD
5 (green) SG 5 SG

Update: A very kind soul in the Los Alamos HPC group has posted scans of ICEBOX documentation: http://institute.lanl.gov/data/linuxnetworxinfo/ Thanks to Christian Ritter for the link.