interesting experience over the last coupla weeks...
generally during peak periods... i'm getting wan "freeze"...
observations;
- isp gateway loses all traffic including arp
- restarting the network ( re-requesting an ip solves it for minimum 6-12mins under really high load aka sat morning / fri night )
- it has presented at really short intervals also... so I don't think the dropping is related to dhcp at all
i had issues with this ISP through the "aussie nbn" access network in the past with dhcp renews being lost in peak times... but they seem to solve it several months ago...
in total during quiet days there might be ~3 drops max... but during a busy night it might drop 30 times... ( had 120 in a 24 hour period over the weekend )...
so... if anyone knows heaps about HFC...
-how could the congestion end up in total loss of traffic?
-why would a dhcp renew fix it ... at least short term... major hint here me thinks...
tommorrow, i'll cycle the modem... and call them... but thought i'd ask about insight... it's like some intermediate access concentrators buffer is overflowing or something???
Whirlpool is littered with threads about HFC uplink speeds being savagely reduced (like less than 1Mbps for a 100/40 plan in extreme cases) for the last couple of weeks. What's your ISP's lease interval? Sounds like multiple renewal requests are getting dropped because of the clogged uplink but I don't know why that would cause session drops unless your router has a low retry limit 
.
1 Like
short aussiebb
"leasetime": 1800
defo was an issue a year ago... but manifested in periods correlating to half the lease times exactly and verified with tcpdumps + non-responses... pretty sure they worked that out... ( either at a nbn qos level or "radius cluster" + upped their lease time used to be 5 mins )
fairly convinced the issue is upstream of my NTD and at or below their gateway... see what they have to say about it tommorrow
tonight there has only been a few... but on the weekend heaps. so i'm leaning on the side of NBN infrastructure being the culprit... not such a big deal for me now tho' ... thankfully most traffic is bufferable... and I am sensitive/understandable...
job interviews tho'... ( all online now )... hopefully won't persist for more than another two weeks... just had me puzzled as to the real issue...
DAMNnnnn! 
Glad i'm not that dude.
More info in relevant ABB thread. May not be your problem but it's causing a lot of aggro with WFH videoconferencing in particular (I'm on FTTN so not affected by this issue but that has it's own issues...).
1 Like