Twitter Service Stopped working
Hi,
This post is more for info, as I have been unable to recreate this.
I was playing around with the twitter feature from my phone. Using ubertwitter to post commands to the relays. I did not get any response. I had got responses in the morning but not in the afternoon.
By response I mean deleting of the command tweet and a tweet from the box, which says "Relay x is off".
When I eventually got home tonight, using the xfx viewer there was no twitter present, in the item list and no heart beat.
I have spent the last while posting various messages and no luck.
If it does happen again, how can I get at information that would be of use to you? Are there any logs etc?
Cheers
Dan
The logs where uber useful I have a fix for this problem (I think). The easiest way to test this is for me just to push a new build and see what happens rather then waiting for twitter to time out its API again, which could take a while. I'll leave my system running twitter with debug on as well.
Unfortunately I've been working on some other changes in the mainline trunk that I have yet to regression test using the older M8 firmware. So if I push this it (might) break everybody's HAH system. I'm using the yet to be released 328 chip and different AVR firmware you see. *sigh* the delimer.
Tell you what I'll push this as a BETA and you can test this out and make sure my other changes are ok too.
/etc_ro_fs/update-dev hah-beta.dbzoo.com
Brett
Well let it run over the weekend before declaring this a success if you don't notice any other defects then I suspect the other changes I did are compatible with your AVR firmware. This is good, they should have been. Once its in the clear I'll push a new build for all.
Brett
I'm going to have to run the twitter in GDB overnight and try and catch the error. It might take me a little to figure out now. *sigh*
... since my last reboot. xap-twitter still running. No tweets handled, but it's there.
Derek.
I ran it in GDB and the program is not crashing per se its actually doing a clean exit. Which is worse a crash is easier to find.
I think there is something funky happening in the Curl library I even ran an strace but it didn't tell me anything I didn't already know.... so now I'm going to have to mock up an apache server that generates a 503 return code and run this snippet locally stepping through the code. As you can tell this is going to take time - rather damn annoying actually.
Brett
I'm pretty sure I've nail this problem update to beta 279.3 let it run for a few days and let me know how you get on.
PS: This build has the new LUA engine so don't just upgrade from the mainline build unless you are prepared to rewrite any LUA scripts you have.
I'll run my twitter in debug mode and see if something turns up - can you run yours in debug mode too?
One of us might catch something useful. I know i fixed the 503 return code issue so not sure what this might be.
I ran my capture for a while but I did not see any issue. I had a power outage and my server rebooted since then I haven't bothered to set it back up again. Having said that I've never seen twitter down since my fix.
If the twitter process isn't showing up in the xfxviewr then it sounds like it has crashed. Unfortunately due to space limitiations on the HAH all core dumps are prevented and even if they where enabled all the binaries have their symbolic debug tables stripped (due to limited flash space). So there is not much you can provide except keeping you eye on it and seeing if its reproducable. I haven't had any other crash reports.
Brett