[Users] problem with all-in-one ovirt install on Centos 6.3 related to hostname resolution
Hello, First, i apologise if it's the wrong place for my question. I had try to install ovirt all-in-one on a Centos6 test server, following this howtos: http://wiki.centos.org/HowTos/oVirt, http://blog.jebpages.com/archives/up-and-running-with-ovirt-3-1-edition/ and http://www.ovirt.org/Quick_Start_Guide. I followed the pre-requisites, cpu, ram, disks, packages etc.. My server is a test server with fresh new Centos 6.3 install, and so, it's in a lan séparated from the main prod lan. But It hass acces to Internet and DNS, but it isn't registred in our DNSs. Simply it's declared in his own hostfile ( /etc/hosts/). It's fqdn is srv-santos.cg33.fr (note the lame joke :-)), and when I ping it from himself (ping srv-santos;cg33.fr), the ping is OK (idem with ping localhost). When with python i do this: gethostbyname('srv-santos.cg33.fr') '172.18.93.215' the response is OK. So the OS himself can resolve is own hostname via the hosts file, since it isn't in DNS. I also checked that in the resolv.conf, the order is hosts, bind. But when i do the engine-set-up i obtain this: srv-santos.cg33.fr did not resolve into an IP address User input failed validation, do you still wish to use it? (yes|no): If i say yes, the install pursue, but it fails: "Error: There's a problem with JBoss service.Check that it's up and rerun setup. Please check log file /var/log/ovirt-engine/engine-setup_2013_03_06_09_52_24.log for more information" The last lines of the log -->File "/usr/share/ovirt-engine/scripts/plugins/all_in_one_100.py", line 216, in waitForJbossUp utils.retry(isHealthPageUp, tries=45, timeout=350, sleep=5) File "/usr/share/ovirt-engine/scripts/common_utils.py", line 929, in retry return func() File "/usr/share/ovirt-engine/scripts/plugins/all_in_one_100.py", line 429, in isHealthPageUp raise Exception(ERROR_JBOSS_STATUS) Exception: Error: There's a problem with JBoss service.Check that it's up and rerun setup. <-- I even started jboss-as during one of my many tries: [root@srv-santos fred]# service jboss-as status jboss-as is running (pid 22043) So I think that the setup process can't resolve the name of the host to test if jboss is up and the servlet ok. This possibly means that the name resolution is only done with DNS, without checking the hostfile. Can someone help me to solve this problem (without seting up a local bind server on my host or registrering my test server in our DNS servers of course:-) Thanx
Hi Frederic, thanks for your report! Il 06/03/2013 10:23, Frederic Bevia ha scritto:
Hello, First, i apologise if it's the wrong place for my question. I had try to install ovirt all-in-one on a Centos6 test server, following this howtos: http://wiki.centos.org/HowTos/oVirt, http://blog.jebpages.com/archives/up-and-running-with-ovirt-3-1-edition/ and http://www.ovirt.org/Quick_Start_Guide. I followed the pre-requisites, cpu, ram, disks, packages etc.. My server is a test server with fresh new Centos 6.3 install, and so, it's in a lan séparated from the main prod lan. But It hass acces to Internet and DNS, but it isn't registred in our DNSs. Simply it's declared in his own hostfile ( /etc/hosts/). It's fqdn is srv-santos.cg33.fr (note the lame joke :-)), and when I ping it from himself (ping srv-santos;cg33.fr), the ping is OK (idem with ping localhost). When with python i do this: gethostbyname('srv-santos.cg33.fr') '172.18.93.215' the response is OK. So the OS himself can resolve is own hostname via the hosts file, since it isn't in DNS. I also checked that in the resolv.conf, the order is hosts, bind. But when i do the engine-set-up i obtain this: srv-santos.cg33.fr did not resolve into an IP address User input failed validation, do you still wish to use it? (yes|no):
This seems ok since it uses nslookup for searching the ip address and it doesn't find the hostname on DNS.
If i say yes, the install pursue, but it fails:
"Error: There's a problem with JBoss service.Check that it's up and rerun setup. Please check log file /var/log/ovirt-engine/engine-setup_2013_03_06_09_52_24.log for more information"
The last lines of the log -->File "/usr/share/ovirt-engine/scripts/plugins/all_in_one_100.py", line 216, in waitForJbossUp utils.retry(isHealthPageUp, tries=45, timeout=350, sleep=5) File "/usr/share/ovirt-engine/scripts/common_utils.py", line 929, in retry return func() File "/usr/share/ovirt-engine/scripts/plugins/all_in_one_100.py", line 429, in isHealthPageUp raise Exception(ERROR_JBOSS_STATUS) Exception: Error: There's a problem with JBoss service.Check that it's up and rerun setup. <-- I even started jboss-as during one of my many tries: [root@srv-santos fred]# service jboss-as status jboss-as is running (pid 22043) So I think that the setup process can't resolve the name of the host to test if jboss is up and the servlet ok. This possibly means that the name resolution is only done with DNS, without checking the hostfile.
I'm using ovirt-3.2 on Fedora 18 and I can't reproduce the JBoss issue using only /etc/hosts without DNS. I don't think it's related to name resolution. AIO uses urllib2 for checking the Healt Page and it should work also if you're using /etc/hosts only. It seems that ovirt-engine service is not started. Please run engine-cleanup, engine-setup and check ovirt-engine status if it fails again. If it fails, can you please attach your logs?
Can someone help me to solve this problem (without seting up a local bind server on my host or registrering my test server in our DNS servers of course:-)
Regards, Sandro
Sandro Bonazzola <sbonazzo@...> writes:
Hi Frederic, thanks for your report! ......;
Please run engine-cleanup, engine-setup and check ovirt-engine status if it fails again. If it fails, can you please attach your logs?
....> Regards, Sandro
Hi Sandro, thanks very much for your answer. I did run engine-cleanup, engine-setup and check ovirt-engine status and it failed again. But, meanwhile, to validate my hypothesis on the faulty name resolution, I installed dnsmasq, a small dns proxy/forwarder, on the same server (srv-santos.cg33.fr), wich can resolve name from local hosts file and from upstream DNS. Then, i re-run engine-setup, and bingo! :-)) no name résolution pbms and no error on JBoss.. Alas! it failed again shortly afterwards with this message: "could not create ovirtsdk API object" It seems to be a known bug cf. "https://bugzilla.redhat.com/show_bug.cgi?id=869457", fixed in lasts fedora's packages.. but apprently not in the ovirt-all-in-one packages on Centos 6. So unless if someone have a solution for this pbm, other than install ovirt on a fedora distrib with the last package of course, i just decided to stay on my centos 6, uninstall ovirt all-in-one and guive a try to a standard engine instal + a standard vdsm install on the same machine with the regular ovirt packages for centos. Regards, Frederic
_______________________________________________ Users mailing list Users <at> ovirt.org http://lists.ovirt.org/mailman/listinfo/users
On 03/06/2013 06:12 PM, Frederic Bevia wrote:
Sandro Bonazzola <sbonazzo@...> writes:
Hi Frederic, thanks for your report! ......;
Please run engine-cleanup, engine-setup and check ovirt-engine status if it fails again. If it fails, can you please attach your logs?
....> Regards, Sandro
Hi Sandro, thanks very much for your answer. I did run engine-cleanup, engine-setup and check ovirt-engine status and it failed again. But, meanwhile, to validate my hypothesis on the faulty name resolution, I installed dnsmasq, a small dns proxy/forwarder, on the same server (srv-santos.cg33.fr), wich can resolve name from local hosts file and from upstream DNS. Then, i re-run engine-setup, and bingo! :-)) no name résolution pbms and no error on JBoss.. Alas! it failed again shortly afterwards with this message: "could not create ovirtsdk API object" It seems to be a known bug cf. "https://bugzilla.redhat.com/show_bug.cgi?id=869457", fixed in lasts fedora's packages.. but apprently not in the ovirt-all-in-one packages on Centos 6. So unless if someone have a solution for this pbm, other than install ovirt on a fedora distrib with the last package of course, i just decided to stay on my centos 6, uninstall ovirt all-in-one and guive a try to a standard engine instal + a standard vdsm install on the same machine with the regular ovirt packages for centos.
Regards, Frederic
Hi Federic, if you just want to have a look at oVirt 3.2, get some experience with it or deploy it. you can try the ovirt-live image[1], just boot it up from a usb and you have oVirt 3.2 up and running. (based on fedora18) [1]http://wiki.ovirt.org/OVirt_Live
_______________________________________________ Users mailing list Users <at> ovirt.org http://lists.ovirt.org/mailman/listinfo/users
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
participants (3)
-
Frederic Bevia -
Moran Goldboim -
Sandro Bonazzola