You are here

Bad Gateway Error-NanoBridge 5G25 on 3.22.8.0

9 posts / 0 new
Last post
KE6GYD
KE6GYD's picture
Bad Gateway Error-NanoBridge 5G25 on 3.22.8.0
Have two NanoBridge 5G25 (32Mb RAM) dishes at a permanent non-critical site on different channels dtd'd to each other pointing to two different hilltops.  They were updated to 3.22.8.0 when it came out and have been working just fine up until a few weeks ago.

Now when accessed via the MESH, would get in Firefox web browser (Also same result in Chrome),  Redirecting then a Bad Gateway error message on both.  Even accessing via the switch at the site caused same.  Downgraded one to 3.22.6.0 and it seemed to work better for a bit.  Then same.

Both Nodes show their name and show up normally as a neighbor when checking MESH Status from the node they connect to.  All Services show up.  One has an LQ 98% and NLQ of 92% with over 50 Mbs.   The other 86% and 88% with 45 Mbs.  

Today, went to the site and downgraded the one again to 3.22.1.0.  That seemed to solve the problem at least for now.  The other one with 3.22.8.0 is still giving the Bad Gateway or is timing out.

Anyone else having an issue with these older devices?  These are the only NanoBridges permanently on our network.  All the others we have are deployed only during drills and events.

Regards
Bob
KE6GYD



 
KM6IAU
KM6IAU's picture
Yes, just had this on a
Yes, just had this on a Powerbeam M5-300.
 

Bad Gateway

The process did not produce any response
w6bi
w6bi's picture
32 MB devices
The devs are pursuing this issue.   Possibly a memory leak; TBD

Orv W6BI
 
KE6GYD
KE6GYD's picture
A reboot makes it work again,
A reboot makes it work again, but then after a number of days of passing traffic, we get the Bad Gateway message. The more traffic passed, the sooner it has the issue.   Had to reboot using SSH as we can't get to the AREDN UI.  

  It's looking like a memory issue based on the data being looked at taken from these 32Mb RAM nodes.
We have confirmed that the issue does not occur with Release 3.22.1.0 (we downgraded the firmware on one of the nodes from 3.22.8.0).  It occurs on Release 3.22.6.0 and continues through the Nightly Build 1787-8fb2869.

So we have two nodes, one working fine with the 1.0, but the other has the nightly and is still exhibiting the issue.

As Orv said, the developers are looking into this.  They're thinking it has something to do with the new code language being used.

Regards
Bob 
 
w6bi
w6bi's picture
Other Stuff
As part of the evolving AREDN software, they're putting more device info in the JSON file that records a node's vital data.  So it's larger, and correspondingly larger on bigger networks.   Older versions of the code didn't have this additional recorded info --> smaller JSON file.
One thing to try is adjust the Memory Thresholds in Advanced Configuration in order to limit the footprint of those large web pages.

If you're going to run snmpd, the ultimate resolution is to retire those 32 MB devices.  Time marches on...

73
Orv W6BI
 
KM6IAU
KM6IAU's picture
PBE-M5-300 is 64MB...
... and not discontinued, yet exhibits the same symptoms.
k1ky
k1ky's picture
Verify AR71xx or ATH79 Bullet Firmware version?
We used to run the ar71xx version and I recently discovered that the ath79 bullet firmware is recommended for the NBM5 units.  Which one are you installing?
KE6GYD
KE6GYD's picture
Mmmm, interesting.  Where did
Mmmm, interesting.  Where did you discover this?  Isn't the chip the determining factor for what to use?  Must have worked for you.

We used the ar71xx since that was only available in the Downloads page for 3.22.8.0 And that's what the hardware is listed as.   But I see there is a nightly ath79 for the bullet. 

Where can I get a copy of the Release 3.22.8.0 ath79 Bullet Sys Upgrade?
 
KE6GYD
KE6GYD's picture
Tried the ath79
Just tried it on a nightly,  but it wouldn't take it.  Must be chip specific.
73
Bob

 

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer