rio-5.0-M3 does not start

[Zsolt Kúti]

I bring this here what was co-mentioned in a different thread.


I try to run 5.0-M3 with jdk7:

$ rio start all
starter config [/home/kuti/apps/rio-5.0-M3/config/start-all.groovy]
Error opening zip file or JAR manifest
missing : /home/kuti/apps/rio-5.0-M3/lib/boot.jar Error occurred during
initialization of VM agent library failed to init: instrument


Indeed, boot.jar is missing!


RIO_HOME is set (double checked).
md5 is as expected.

Zsolt

[Dennis Reedy]

Zsolt,

My guess is you have RIO_HOME set to an older version of Rio. I just downloaded 5.0-M3 and starts for me just fine.

HTH

Dennis

> --
> You received this message because you are subscribed to the Google Groups "Rio Users Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to rio-users+...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Zsolt Kúti]

On Mon, 8 Sep 2014 10:22:11 -0400
Dennis Reedy <dennis...@gmail.com> wrote:


> Zsolt,

> My guess is you have RIO_HOME set to an older version of Rio. I just
> downloaded 5.0-M3 and starts for me just fine.

- Tho experiment with this version in a shell I use "export RIO_HOME=...
5.0-M3"
- echo $RIO_HOME show this happened
- if I start the script from what any other shell opened that would have
inherited this environment variable, right? I cannot imagine how after
setting that variable could change the value causing what you suspect.

But I will restart my machine after permanently setting the var and
let's see if that changes the situation.

Thanks
Zsolt

[Dawid Loubser]

If your RIO_HOME is normally set by some persistent environmental
setting (such as in a .bashrc, or /etc/profile.d/*.sh file), setting the
variable in a terminal, and then opening a new terminal, will not result
in the new terminal having your overridden value.

Best check the value after you've opened a new terminal.

regards,
Dawid

[Zsolt Kúti]

On Mon, 8 Sep 2014 16:42:19 +0200
Dawid Loubser <dawid....@gmail.com> wrote:


> If your RIO_HOME is normally set by some persistent environmental

> setting (such as in a .bashrc, or /etc/profile.d/*.sh file), setting
> the
> variable in a terminal, and then opening a new terminal, will not
> result
> in the new terminal having your overridden value.
> Best check the value after you've opened a new terminal.

Thanks David, I am aware of this and does everything in the same
terminal. The best I can think of now to set it permanently, restart
the machine and see what happens. I'll shortly do that.

Zsolt

[Zsolt Kúti]

It seems the RIO_HOME exported though, was not inherited. Its permanent
setting took me one step ahead.

Now Aether wants to download necessary dependencies from
repo1.maven.org and fails. The usual OS level proxy settings
(permanently set http_proxy/ftp_proxy env vars) are not used.
Still searching for a solution.

Thanks
Zsolt

[Dennis Reedy]

This only happens if Rio cannot install itself into your local Maven repository. This is still pointing to an environment issue IMO. Take a look at the install.groovy script, or try running it from the command line.

Dennis

[Zsolt Kúti]

On Mon, 8 Sep 2014 11:19:41 -0400
Dennis Reedy <dennis...@gmail.com> wrote:


> > Now Aether wants to download necessary dependencies from
> > repo1.maven.org and fails. The usual OS level proxy settings
> > (permanently set http_proxy/ftp_proxy env vars) are not used.
> > Still searching for a solution.


> This only happens if Rio cannot install itself into your local Maven
> repository. This is still pointing to an environment issue IMO. Take
> a look at the install.groovy script, or try running it from the
> command line.

It turned out* that on Windows and Gnome 2 systems java can get proxies
from the desktop environment. The rest of the world has to set them
manually either globally in <openjdk7>/jre/lib/net.properties or for the
jvm to be just run.

* http://docs.oracle.com/javase/7/docs/technotes/guides/net/proxies.html


Now I could start rio (Sigar is put aside).
The log shows cybernode is started:
INFO 08:54:00.311 o.rioproject.cybernode.CybernodeImpl - Enlist with
ProvisionManagers as an instantiation resource using:
Cybernode-2279@kuti-zsolt
DEBUG 08:54:00.317 o.rioproject.cybernode.CybernodeImpl - Created new
ServiceID: 8d34ca40-40d0-4b59-b3c2-133901709582
INFO 08:54:00.317 o.rioproject.cybernode.CybernodeImpl - Started
Cybernode [kuti]


but when it comes to provisioning services it claims:
"There are no registered Cybernodes" and services are not provisioned.
rio-ui shows no cybernodes, no services (only an orange diamond).
However, rio-ui from 4.2 distribution still shows cybernodes running on
remote machines.

08:34:29.629 org.rioproject.monitor.services -
ServiceElementManager.verify(): [Application/Outrigger] actual [0],
pending [0], maintain [1]
WARN 08:34:29.630 o.r.monitor.ServiceProvisioner - Failure Reason(s):
There are no registered Cybernodes, unable to provision Outrigger
DEBUG 08:34:29.630 o.r.monitor.ServiceProvisioner - Wrote
[Application/Outrigger] to Dynamic-Service TestManager
DEBUG 08:40:46.682 o.r.cybernode.ServiceConsumer - Canceling all event
registrations to Provisioner instances


Until now the same opstring, deployment type (everything under deploy
dir) and unchanged service code is being used to proceed gradually.

Zsolt

[Dennis Reedy]

So the Cybernode cannot join the network it seems. Lets think about what has changed ... new operating system and an upgrade to Java 7. Can you please try and start the Cybernode using the cybernode script (not startall), and also start the Provision Monitor in it's own vm using the monitor script. But turn on logging for net.jini.discovery to Level.ALL?

Dennis

[Zsolt Kúti]

To narrow the problem I simplified my tests and started rio by leaving
out my services for the time being.
Comparing the logs the difference is the below messages that are
present on Ubuntu, missing on FreeBSD:


DEBUG 12:04:07.238 o.r.cybernode.ServiceConsumer - ProvisionManager
discovered
org.rioproject.monitor.ProvisionMonitorProxy$ConstrainableProvisionMonitorProxy@6f81d54b
DEBUG 12:04:07.817 o.r.monitor.ServiceProvisioner - Registered new
Cybernode-7375@molnar-szilvia, count [1] DEBUG 12:04:07.834
o.r.cybernode.ServiceConsumer - Established ProvisionManager
registration INFO 12:04:07.836 o.r.cybernode.ServiceConsumer -
Registered to a ProvisionManager, now connected to [1] ProvisionMonitor
instances


Summing it up, for some reason it seems cybernode is created but not
registered, possibly because no ProvisionManager is discovered on
FreeBSD.
Add to it the further observation that rio-ui is not showing existing
cybernodes, while on Ubuntu it is. As if discovery did not work on my
machine?

Does that make sense?

Zsolt

[Zsolt Kúti]

Just arrived, when had sent mine...
I am doing it.

[Zsolt Kúti]

On Tue, 9 Sep 2014 08:06:22 -0400
Dennis Reedy <dennis...@gmail.com> wrote:


> So the Cybernode cannot join the network it seems. Lets think about
> what has changed ... new operating system and an upgrade to Java 7.
> Can you please try and start the Cybernode using the cybernode script
> (not startall), and also start the Provision Monitor in it's own vm
> using the monitor script. But turn on logging for net.jini.discovery
> to Level.ALL?

To make it short it was a FreeBSD (java) specfic thing.

And for the archive: from a corner of my memory I recollected that
there was (and seemingly still is) a problem around ipv4/6 code working
together. Most of the time this results in some visible error
message, here is not:
Using -Djava.net.preferIPv4Stack=true allows me to what I want.
Now I can step to my services...

I still do not now why came this up on my new system and not on the old
one.

Anyway, thank you for your help and patience.

Zsolt