# ### It is now Mon Dec 15, 2014 ### # # -> (14:04) From prox [@work], to FTTH, cxn: # - Hmph. My parents keep having FiOS outages at home. They happen about # - once a month and are confirmed (with Verizon). Unfortunately after every # - outage their router (Linux box) gets a new IP in a completely different # - netblock. They usually either bounce the Linux box or wait until DHCP to # - refresh itself. They are thinking of going back to Cablevision (and # - paying more). This seems out of the ordinary, right? Have other folks # - here had similar issues? # # -> (14:06) From keith, to FTTH: # - I used to get new IP addresses from VZ fairly regularly. This has slowed # - down quite a bit in the last year or so. # # -> (14:12) From prox [@work], to FTTH: # - What's slowed down mean now? Every month or every 4-6 months? # # -> (14:12) From prox [@work], to FTTH: # - I'm thinking of hacking dhclient to re-request the lease every 5 minutes.. # # -> (14:14) From keith, to FTTH: # - I can't remember the last time it changed. At least 6 months. # # -> (14:14) From prox [@work], to FTTH: # - Alright, thanks. # # -> (14:15) From keith, to FTTH: # - Before that it was more like a monthly thing at best. # # -> (14:30) From FliesLikeABrick [@work], to FTTH, cxn: # - my experience with Verizon was that a given broadcast domain had a large # - number of small (/24) DHCP pools, with a 1 hour TTL, and no real attempted # - re-use of your "last lease" past the expiration of the lease. The result # - of this being that if your DHCP client/CPE was offline 1 second past the # - end of the lease or explicitly released the lease, you got a different # - address in a different block every time you got back on # # -> (14:31) From FliesLikeABrick [@work], to FTTH, cxn: # - but if your 1 DHCP client stayed online for months or years with no # - outages, you continually got the same IP address reliably # # -> (14:31) From FliesLikeABrick [@work], to FTTH, cxn: # - that's a long-winded way of saying "addresses weren't changing willy-nilly, # - but there were not measures in place to prevent address changes outside of # - the guaranteed lease. renews kept the same address, but requesting a new # - lease post-expiration gave you a new one"... it's pretty typical behavior # - but their relatively short lease time made it stand out # # -> (14:34) From prox [@work], to FTTH, cxn: # - Ah, that's helpful information. So, the IP changes are just a result of # - the outages and their host being offline for awhile. Sigh.. now to find # - out why there are so many outages. # # -> (14:35) From FliesLikeABrick [@work], to FTTH, cxn: # - correct. while the behavior is not ideal for someone who wants to run a # - service at home, it's nothing abnormal/bad # # -> (14:36) From FliesLikeABrick [@work], to FTTH, cxn: # - (it's not like the address is changing on every renewal and causing all # - existing connections to time out). This is why someone above said that # - they've the same address for 6 months or a year- they just have particularl # - y stable power and no verizon-side outages past an expiration time # # -> (14:36) From prox [@work], to FTTH, cxn: # - Hm, I'm seeing " option dhcp-lease-time 7200" - so that 1 hour TTL. The # - problem is, the Linux box does not know to try to get a new lease. I'm # - going to see if I can override that option in dhclient.conf. # # -> (14:36) From prox [@work], to FTTH, cxn: # - (that 7200 came in with the DHCPOFFER) # # -> (14:37) From FliesLikeABrick [@work], to FTTH, cxn: # - it's possible I'm remembering the lase half-life, or that verizon doubled # - their lease time (or it varies slightly from market to market) # # -> (14:37) From FliesLikeABrick [@work], to FTTH, cxn: # - lease* # # -> (14:38) From prox [@work], to FTTH, cxn: # - Hmm, I can set "dhcp-lease-time" in dhclient.conf, but I have a feeling # - that's only the desired lease time sent to the DHCP server. I guess they # - can come back with anything. Let me try.. # # -> (14:39) From FliesLikeABrick [@work], to FTTH, cxn: # - in any case, trying to get a longer lease time is just putting a band-aid # - on the underlying issue (the frequent outages) # # -> (14:40) From prox [@work], to FTTH, cxn: # - I'm actually lowering it. I want dhclient to renew the lease every 5 # - minutes so my parents don't have to wait an hour or two for their Internet # - to start working again after the "outage" has passed. # # -> (14:41) From FliesLikeABrick [@work], to FTTH, cxn: # - aha, now I understand. I suspect that won't work # # -> (14:41) From prox [@work], to FTTH, cxn: # - Odd.. I set it to 300 but dhclient says "bound to 96.242.82.14 -- renewal # - in 2880 seconds." - the problem is it still says 7200 in the lease file. # - Hmph. I wonder if I just need a cronjob :( # # -> (14:41) From FliesLikeABrick [@work], to FTTH, cxn: # - once the client has a lease, it's supposed to try to renew at 50% and # - 7/8ths time to expiration. I suspect that once the client has a lease, # - its attempted renewal is based on that. You could fake it by running a # - ... yep, cronjob to run dhclient # # -> (14:42) From FliesLikeABrick [@work], to FTTH, cxn: # - re-running dhclient won't re-up the lease. for example I just re-ran # - dhclient on a machine at home and it told me 'renewal in 120608 seconds' # - which is obviously not a nice round number # # -> (14:43) From FliesLikeABrick [@work], to FTTH, cxn: # - so 2880 seconds means that you probably last "actually" renewed your lease # - 720 seconds before that # # -> (14:44) From prox [@work], to FTTH: # - ,-cxn;I don't want it to re-up the lease, I just want it to try more often. # # -> (14:44) From FliesLikeABrick [@work], to FTTH, cxn: # - (because renewals on a 7200 second lease will be attempted at 3600s, and # - 3600-2880=720) # # -> (14:44) From FliesLikeABrick [@work], to FTTH, cxn: # - right, I understand that -- I'm just trying to add color to your comment # - about it being odd that you got 2880 seconds in the output from dhclient # - despite the lease time being 7200 in the ifle # # -> (14:44) From FliesLikeABrick [@work], to FTTH, cxn: # - file* # # -> (14:44) From prox [@work], to FTTH, cxn: # - Ahh, ok. # # -> (14:45) From c, to FTTH, cxn: # - (cron job) I had to do that with rr I think it was a decade ago (or was it # - RCN/Comcast at my parents?). It meant the machine would re-requst a lease # - without me logging in; otherwise sometimes it would just stop working # - until I sshed in on the LAN to the openwrt device. # # -> (14:45) From prox [@work], to FTTH, cxn: # - Got it. "supersede dhcp-lease-time 600" - renews ever 5 minutes, in # - theory. Lease DB reflects the change. # # -> (14:49) From prox [@work], to FTTH, cxn: # - Yep, tcpdump confirms behavior. Cool. Now, I wish they wouldn't have # - outages. (End of review)