notes.txt
author stephen
Thu, 17 Aug 2017 12:43:07 +0100
changeset 62 5fa8703565c4
parent 53 72d901089cb2
child 63 5c92de4caca6
permissions -rw-r--r--
aug17
     1 
     2 20170111 - device online for 1 min @0948 - someone on roof?
     3 20170112 - device offline at 1524, surprised if that's power? check
     4 20170113 - device work @ 0933 briefly - see early-wake-theory below
     5            early-wake-theory:
     6             if device goes below 11.1V while sleeping then it won't wake on
     7             the RTC but will wake when the solar charger sees 12V again,
     8             at which point it will see that it ought be asleep unti 1100
     9             and so it dozes until then - possibly means that we're keeping
    10             the kerlink up after the eurotech has gone to sleep, otherwise
    11             why is the voltage still dropping?
    12             - I modified /etc/stopproc.sh to stop pbm and call ker-off
    13 20170124 - took that out again, added code to pbm binary
    14            looks like that fix didn't do it, need to test
    15            some more when at node Monday
    16 20170118 - fixed kerlink not powering off - both s/w and wiring!
    17            we re-charged unit from mains around 1400-1430, after
    18            that back to solar, hopefully voltage on 19th will
    19            behave
    20 20170124 - validated that ker-off does remove voltage from kerlink
    21            power input; found a code path in pbm that wasn't 
    22            calling ker-off, testing of that needs to be finished
    23            when device has power again. Note: left device with
    24            sleeping only set from 1440 to 1455 so it'll wake 
    25            early on 25th (not likely to get more power today
    26            due to sun angle)
    27 20170125 - eek, I'd put the ker-off code in the wrong place;-)
    28            fixed now I think finally
    29 20170126 - seems to have worked out - node went down @1600
    30            on 25th with battery @12.52V, awoke on schedule 
    31            @1102 on 26th with battery @12.49V and weather was
    32            overcast all morning (that I saw:-) on the 26th,
    33            node is drawing about 0.9-1.1A around 11am
    34            probably only safe to fully believe stats from
    35            here on, though some info can be gleaned from
    36            earlier data no doubt
    37 20170127 - there was a wake @ 10:02 - might be ok though, node
    38            went into long standby at 1459 on the 26th so was
    39            close to out of power (11.34V)
    40 20170128 - looks like basil was rebooted about 17 hours ago 
    41            (around 1800 on the 27th?) (down is also down)
    42            so vpn wasn't up - not sure if client will connect
    43            before it's next reboot, hasn't so far after 5
    44            mins (vpn manually restarted @ 1140) I'm guessing
    45            there was a power off event
    46 20170201 - basil reboot: there was a bridge0 that was ifup'd
    47            and that might be interfering somehow, took that
    48            down, we'll see @ 1100 if that helps, if not, I'll
    49            wanna go check unit physically (also could be 
    50            SIM card crap, check that too - SIM seems ok, has
    51            7GB and 8 days left) 
    52            Turns out that the SIM was not ok - the "add on"
    53            thing is not being used and the balance was down
    54            to 0.65c or actually maybe negative. Topped the
    55            fecker up.
    56            Clock was still on jan 29 today, odd - maybe
    57            rtc isn't working right? also that might interact
    58            with vpn, not sure - actually no, it was that
    59            the node thought it was 1600 on the 29th so it
    60            went to sleep, wonder when it'd have awoken if
    61            Kerry hadn't hit the physical switch?
    62            I have the kerlink boottimes.txt that could be
    63            used to fix the boot times and hence battery
    64            log of the loradtn node if we want
    65 20170202 - odd timing - node woke @1047 - is power still being
    66            consumed?
    67            also odd vpn pattern - new client IPs every few
    68            minutes, check node's syslog later
    69 20170206 - looks nominal at the moment, nice weekend weather
    70            helps:-)
    71 20170213 - VPN went down at 1427 in good bright weather with
    72            battery nicely high  (13.28V at 1357), GSM balance
    73            also fine, but maybe it's the add-on fecking thing
    74            I added 3GB of data (good 'till March 15) which'll
    75            take effect at next connect (reboot?) it didn't
    76            reconnect in 5 mins anyway. Or, it could just be
    77            a DHCP lease meets VPN issue. Check tomorrow.
    78 20170214 - yep, yesterday's outage was gsm related - pppd saw
    79            a disconnection. Device stayed powered up 'till
    80            1600 as planned, but had no n/w. Once it powered
    81            up today, it was back on line. Must check that
    82            again on March 10th (set a reminder:-)
    83            Meanwhile no sun today, so node will likely go
    84            to sleep in an hour or so (now @12.27V and drawing
    85            0.9A) 
    86 20170222 - seems like there is some browning out happening,
    87            today and on the 18th - might be the higher load
    88            means that the thresholds aren't right for this
    89            setup with voltage is marginal (e.g. 11.5V)
    90 20170228 - graphs show node up yesterday but I didn't see 
    91            that from CLI, check it out
    92            also - redo scripts to add March
    93 20170308 - we added the code for the 2nd ammeter for the 
    94            device under test (DUT). Noted that the voltage
    95            calibration may need a bit of work, but not
    96            enough to make a change now as that'd invalidate
    97            data since Jan. Worth doing more before any 
    98            further deployment. The new pbm code was tested
    99            on the bench and deployed to the rooftop node
   100            at 1500 today. We'll follow up with the new
   101            phidgets/veroboard assembly shortly (so DUT
   102            figures for now will be garbage) some voltages
   103            seen on the bench power supply and as seen by
   104            pbmd are:
   105               Bench                 pbmd
   106               11.6                  11.87
   107               11.8                  12.10
   108               12.8                  12.97
   109               13.2                  13.44
   110            Current measurements OTOH, seem accurate 
   111            enough to within 100mA or so.
   112 20170310 - veroboard with new ammeter deployed today (so 
   113            some outage this morning) DUT readings showing
   114            about 300+/-40mA
   115 20170317 - added a graphic showing the DUT power consumption
   116            vs. the overall power log
   117 20170321 - extended sripts to cover to end-April
   118            also noticed that SPIKE readings are being generated
   119            when voltage is identical in two successive readings
   120            probably a pbm bug, but not a biggie
   121 20170408 - topped up SIM card - balance was small and device
   122            offline on a sunny day, see if fixed with tomorrow's
   123            reboot
   124 20170409 - seems like that's it, up nomimally this am, and
   125            past three days readings seem corrupted (no NTP to
   126            set date, odd that it falls back to the last "good"
   127            day, something to check out)
   128 20170416 - seems like the root FS is full on DTN node, note
   129            sure since when, cleaning today. Also can't ping
   130            the kerlink, not sure if connected. Need a way 
   131            for unattended upgrades to not fill / based on
   132            adding stuff to /boot. Fixed that manually and
   133            did a reboot at 1500 IST... unit came back up
   134            fine, but still no sign of kerlink. Will have to
   135            go over and see Tuesday
   136            oops - node went offline in the middle of an
   137            apt get upgrade about 1530 local, not sure if 
   138            power (should be enough, I'd have thought) maybe
   139            vpn, check tomorrow
   140 20170417 - further checking... unit came up @1100 nominal
   141            last time kerlink asked for dhcp was march 25
   142            finish update for dtn node then check dhcp is
   143            running (looks like it might not be)
   144            Ah-ha! kerlink has moved to .17 from .11 for
   145            some reason, and syslog I copied down yesterday
   146            seems odd, more when endlless update done
   147 20170418 - nominal, still not sure what's up with syslog
   148            but will look tomorrow
   149 20170516 - refreshed vodafone a/c, was offline for a week
   150            or two before that, back now
   151 
   152 
   153