StarV3 v1.1.10 build 1734 has been released.

[tony]

The new release can be downloaded from the http://www.star-os.com (http://www.star-os.com/) downloads page.
This is an important update for those using releases v1.1.6 through v1.1.8

[Beebe]

Something funny is going on again with the OSPF. It was working in version 1.1.4 and I upgraded to version 1.1.10 and OSPF would restart itself every few seconds.

I restored factory defaults on OSPF and then it would keep running, but when I reconfigured it i could add all the configs except specifying the network. Whenever I do that OSPF restarts itself.

Thanks,
Roger

[tony]

That is interesting. Can you provide us with the command in question that is causing it to restart? We will do some investigation.

[Beebe]

In OSPF I can do

router ospf
ospf router-id 12.109.193.1
neighbor 12.109.193.2

and it runs ok after these commands in as far as it doesn't restart, but when I do

network 12.109.193.0/24 area 0.0.0.0

Then a few seconds later I will be kicked out of the OSPF configuration screen, and back to the main GUI. It shows OSPF as not running but it immediately starts running again and I can go back into it and the network statement is gone. If I do the network command and then save changes really quick before it boots me out then ospf restarts over and over.

Thanks,
Roger

[tony]

Thank you, I will do some digging into this.

[David L. Vrablic]

I have been working with a very poor path on 5.x.
It is across 2 city blocks to a rooftop with a treeline with denuded branches. I can see the antennas through the brush from the pole location.
WAR4 boards w CM-9 both ends and it really seems like it should work just fine.
In order to get it to associate at all I even changed antennas on both ends and have a big ugly 24 Dbi patch on the pole and a grid 24 Dbi on top of a school turning this box into a point to point link.
-----------
Funny thing is I have very little improvement over the origional 12 Dbi omni on the school and a 19 Dbi patch at the pole.
--------
Association was sporadic at best with a -84 signal.
I upgraded both ends from 1.1.1 to 1.1.10 to see if that would help.
I did not see any noteworthy improvement, (I just felt better that I now have the latest FW.)
I was finally able to get it passing traffic when I changed to 4x mode at both ends.
The best quality I can get is
Q % Sig Ack Rx Tx XT (No traffic at the moment)
100 -81 -80 6 6 C (Set to 4X mode)
I have come to the conclusion that 5.x is going to have to be truly LOS at any distance to be reliable.
The only thing left is to try Horiz pol but that won't help much when the leaves come back.
Maybe a cloaked 2.4 will make this happen
--------------
Next Subject: I am trying to figure out the site survey .

It seems that if I enter ANY in the essid and do a scan I get several cells information.
If I enter the Essid of one of the sites and get an association, none of the others sites that I saw previously are displayed.
Only the one I am attached to is there.
------
I am not sure but I sware a couple of times I saw one other as well but the testing was fast and furious yesterday, up in the bucket in the pouring drizzle of cold rain.
----
It almost looks like there is some kind of mimimum signal threshold before the site is displayed ?
Maybe an internal code that only displays the data for a site once associated ?
Could someone please shed some light on this subject as it is all I have to work with when setting up new sites.
------------
Merry Christmas from our house to your house.

[lonnie]

Site surveys rarely show other systems on the first scan. Do a second one about 20 seconds later and you'll get lots of sites. There is no signal threshold and you'll see units reporting with -94 dB signals.

Be aware that many schools use old Tsunamis and other FDX full time gear and they take the WHOLE band for 10 mbps of throughput. You'll find it real tough to operate if you have such a system close by.

With such a short shot you can likely use 2.4 GHz and get 15 to 20 mbps sustained. Keep the antennas large to have a small beamwidth so you do not hear everything else and maybe turn power down so you have -60 dB signals.

[David L. Vrablic]

Thanks for the survey info.
I just never seem to observe it working the same way twice.
I will sit down in a controlled enviroment and see what I can learn .
-----------
This old school is now a Boys Club and is so poor "If they could get around the world on a nickle, they couldn't even get out of sight."
They are lucky to have a computer for the office.
It seems that this is where the cameras go, to protect the poor folk and take back the neighborhoods from the drug dealers.
I am not saying the neighborhood is tough but Erik was going to buy me a flack jacket for Christmas.!
--------------
I am going to need 3 cameras on this leg so it will take about 6 meg.
I should be able to get that easy.
I will change out the panel for a 2.4 change the antenna on the other end and give that a try. Gotta love using CM-9's in every box.
Change directions, modes, freqs on a dime.
Thanks and have a great weekend.

Added: I forgot to thank Tog for his off line response as well.
It is nice to double check my instincts with others.

[sligbot]

We've noticed that 5.x gear absolutely requires LOS. Anything in the way will kill the signal and fast. You should easily be able to setup the link with 2.4 gear instead and the lovely cloaked channels do amazing things for interference. We're actually talking about the cloaked channnels as our "secret weapon" in our area because of the ability to jump in between the used channels. If you get a 10MHz channel working with "G" mode, you should easily be able to support a nice 6mbps link .

[soulmata]

Given the beamwidth and absorption rate of 5GHz frequencies, it's pretty much a given that you require LoS. I'm not sure why anyone would even try non-LOS with such a high frequency. It's one of the reason home 802.11a access points have failed so much. Remember that the higher your frequency, the easier it is absorbed and the narrow its beamwidth is.

[go.fast]

I'm not advocating using 5 gig to go throuh trees, but I have a couple shots that have trees in the fresnel zone and they work great.

This stuff is so much easier than the original trango with dish I started out pointing. I remeber that it was time consuming pointing those guys Where these guys are almost as easy as 2 gig shots,

[David L. Vrablic]

I hear wat you are saying.
This path really fooled me.
I can see the other box and antenna from the rooftop.
What was so strange was how little changing from a omni on one end and a panel on the other---to a grid on one end and a big panel on the other end made.
Only about 2Db or so.
Those must be Magic RF snatching branches in the F zone that just suck it all up before it get to the other end.
Also there might be something actually wrong on one end or the other.
It was just to wet and cold to do much.
Tomorrow I am adding the 2.4 G to the mix.
Lets see what that does.
At least the antenna bracket is the same for both antennas.

[Beebe]

Thank you, I will do some digging into this.

Hi Tony,
I have some more information, I've just noticed that the 1.1.10 PC version runs OSPF without problems, but I just tried another WAR board with 1.1.10 to make sure of what I was seeing and it behaved just like the first one I tried. I rolled it back to 1.1.9 and it solved the problem. So the problem appears to be specific to the WAR (I haven't tried it on a wrap).

Thanks,
Roger

[tony]

Thank you. The 1.1.10 (and BETA 1.2.0b) both use Quagga 0.99.5 (latest), however 1.1.9 and previous 1.1.x releases use Quagga 0.99.4.

If the Quagga group has not addressed this problem in 0.99.5 yet, we will roll it back to 0.99.4 for the next release.

Thanks!

[nickwhite]

To add to this and my other post in the beta thread Tony, I did try upgrading that system to 1.1.10 (instead of beta 1.2.0; it was previously 1.1.4) and it exhibited the same behavior that I posted in the beta thread - quagga/OSPF reboots.

[mimbach]

I saw the exact same problem with it crashing again and again. I went through all my configs on all my war boards, added some additional config lines like; ospf priority. After I redid everything my rebooting went away.
So to let everyone hear this, I had the same problem. When I added all the recommended config lines to my radios the problem went away of quagga rebooting constantly.
Please keep in mind guys anytime I saw more then one "DR" listed in the "show ip ospf neighbor" command output is when I ended up seeing problems. Fixing that made my problems go away.

We run ospf on all of our towers now with a total of 18 war boards used for backhauls, a loop for redundancy, and canopy to feed the customers.

[nickwhite]

I saw the exact same problem with it crashing again and again. I went through all my configs on all my war boards, added some additional config lines like; ospf priority. After I redid everything my rebooting went away.
So to let everyone hear this, I had the same problem. When I added all the recommended config lines to my radios the problem went away of quagga rebooting constantly.
Please keep in mind guys anytime I saw more then one "DR" listed in the "show ip ospf neighbor" command output is when I ended up seeing problems. Fixing that made my problems go away.

We run ospf on all of our towers now with a total of 18 war boards used for backhauls, a loop for redundancy, and canopy to feed the customers.

Can you give any more details or examples on the configuration that you cleaned up or "recommended config lines"?

[mimbach]

Guys I think the best thing we can do is post our network layout and ospf configs. Then discuss them as a group. Then everyone can learn and get around this problem. My network is larger with alot of redundant paths and different ospf areas. Thus probly not the best to be an example network.
If you post your network layout and configs I will help and even re-post the configs how and why I would do them.

[Beebe]

I can confirm this behavior, if you specify the ip ospf priority it does run without rebooting. Thanks for the tip!

More info on ospf priority...
http://www.avici.com/documentation/HTMLDocs/02223-06_revBA/ospf13.html

Maybe Tony should continue using the newest version in future releases?

Thanks,
Roger

[tony]

Thank you for the update. Seeing as there is a solution, we will continue using the newer quagga releases unless something more serious pops up.

[nickwhite]

Awesome... I'll make sure to do this now. Thanks guys!

-Nick

[Beebe]

I recently set up a server for a firewall/bandwidth shaper between my T1 router and my switch at the NOC. It was running 10.1.10 on a 1ghz AMD processor. It would reboot about once a day, and one time it rebooted over and over several times.

So I took a second server, loaded 10.1.10 onto a flash disk and put it in a flash to IDE converter. I backed up the config from the original server, and restored to the new box (actually an older 500mhz box). It also reboots about once a day.

Any chance someone can give me a download URL for an older version so I can see if it's a problem specific to this version? It doesn't seem to be available on the download server.

Thanks,
Roger

[tog]

Sorry, I don't have any of the non-WRAP x86 edition of v3. Only the XSCALE and WRAP images.

[greg]

I have a bunch of x86 versions saved. You interested in posting them Tog?

[tog]

I don't know, StarV3 is not an EOL product... so I don't know about posting old versions of it permanently like I have already done with v2.

Only if Valemount doesn't mind, I could post 1.0.3 and 1.1.4 there.

For now maybe you could just PM Beebe and provide them privately to help Beebe out?

[lonnie]

We don't mind if anybody hosts the archive. In order to have people use only the newest we will only post the newest code, but we will not obhect if they have an archive of older code. Older code is not supported and we will only troubleshoot new images. In the past we have worked on older releases and troubles only to find out that a new kernel or driver or whatever had already fixed the issue.

[Beebe]

Last night it went to rebooting over and over... so I logged into it and turned off snmp and it's been ok since. I don't know if that solved it, or if it will start rebooting again, but I'll leave it like it is for now.

Thanks guys,
Roger

[tony]

Please keep us informed. While net-snmp should behave, there is a chance it could be causing some of the problems you are seeing.

[Beebe]

Well it just rebooted again so it wasn't the snmp doing it. So it looks like I'll be needing an older version after all.

Thanks,
Roger

[David L. Vrablic]

Howdy folks,
I have a few units with 1.1.1 out there.
A few with 1.1.4 and most with 1.1.10.
In an attempt to get everything playing from the same page:
Are there any problems bringing them all up to 1.1.10. ????
I am not using any special functions or routing.
Everything is pretty plain point to point and p t mp routed links.
-----------------
One thing has been observed:
When a new unit has been brought online and the association verified from the AP side....
I have to ping the gateway (Or beyond) from the client to get an indication of traffic flow at the AP.
Just connecting a camera doesn't do it.
I would have thought that there would be some data flow at connection.
Any thoughts?
-----------
Gosh it is cold out there today.

[lonnie]

We have our AP at a mix of 1.1.4 and 1.1.10 with no difference in operation or performance. We do have snmp shut down since we feel there is a memory leak and are working to find it in non operational AP units.

As to the camera, what is it talking to that would show traffic on the association screen? Most cameras I have seen do not speak unless spoken to.

[tog]

Using WDS? If using WDS, in 1.1.x right now the client has to send out something first in order for you to be able to communicate with it.

[tony]

That is correct. Since WDS connections are dynamic, they have to be initiated by the client, during it's first data transmission. If your IP camera does not transmit traffic, the WDS connection may never get established.

What some have done, is enable STP on the clients (though not on the AP), which will activate the WDS connection immediately.

[David L. Vrablic]

Thanks everyone,
Nope, not using WDS for this project.
That inf was what I needed to hear.
I am gpoing to standerize on 1.1.10 for the moment.
Everything is going well and expanding like crazy.
Now if I can only get some hardware in house.

[mimbach]

We have had good luck with most of our ospf links.
We have found some areas that are larger with more links and multiple paths that we are not able to make work properly. I have spent 5+ hours reading all the quagga mailing lists and was not able to see any mention of the problem we are seeing. We are seeing ospf die and then the watchdog restart it. When this happens ospf does not always work and we actually just end up rebooting the radio to get our routes to come back up.
Lonnie and Tony, I have tried my best, let me know what you need from me so we can figure this out.
Thank you.

[lonnie]

We are about ready to release a new image with the very latest quagga. It has lots of ospf fixes in its changelog so hopefully they get it right.

[mimbach]

Thank you, I will be waiting anxiously.