Report information
The Basics
Id:
45292
Status:
resolved
Estimated:
48 hours (2,880 minutes)
Worked:
48 hours (2,880 minutes)
Users:
tmark: 48 hours (2,880 minutes)
Priority:
Low/Low
Queue:

People
Requestors:
Cc:
AdminCc:

BugTracker
Version Fixed:
4.4.0
Version Found:
(no value)
Versions Affected:
(no value)
Versions Planned:
4.4.0
Priority:
P1 High
Severity:
S2 Normal
CVSS Score:
(no value)
CVE ID:
(no value)
Component:
(no value)
Area:
feature

Dates
Created:Thu, 25 May 2017 12:10:48 -0400
Updated:Tue, 12 Dec 2017 05:54:17 -0500
Closed:Tue, 12 Dec 2017 05:54:17 -0500



This bug tracker is no longer active.

Please go to our Gitlab to submit issues (both feature requests and bug reports) for active projects maintained by Internet Systems Consortium (ISC).

Due to security and confidentiality requirements, full access is limited to the primary maintainers.

Subject: [BlueCat] dhcp-cache-threshold for dhcpv6
In Support #11511, BlueCat have asked us if we can implement a dhcp-cache-threshold feature for v6 that works more or less the way the existing feature works for v4. > Hello, > > The man pages for dhcpd.conf state that the 'dhcp-cache-threshold' > option is only available for v4. > > Is there any reason for not implementing this option for v6? We > are observing misbehaving dhcpv6 clients in a customer's environment > that are renewing at a rate of almost 1 per second, which causes > an impact in our integrated IPAM solution as we relay those > notifications back to our management system. > > Fernando
I reviewed the code: - I pushed a few cosmetics stuff, please pull and review them - I have a question which IMHO requires an answer: when T1 and T2 are not zero or derived from lifetimes I think they should be reduced? Note this is really specific to DHCPv6 as for DHCPv4 this is only one timer doing everything. My concern is the the threshold is a bit high after a reuse the returned T1/T2 can (mis)lead too far.
We have some feedback from testing of the pre-release patch. #dhcp6s.conf authoritative; ddns-update-style interim; default-lease-time 86400; dont-use-fsync true; subnet6 FC00:12::/64 { client-updates false; ddns-domainname "test.com"; ddns-updates true; do-reverse-updates true; range6 FC00:12::1 FC00:12::1000; } subnet6 FC00:17::/64 { range6 FC00:17::1 FC00:17::400; } There is no failover, and leases are being allocated from the configured pools - no static or reserved involved. The client is running Ubuntu and DHCP Client 4.3.5 The tests included this one: > - V6 leases cannot be reused if DNS information changes, so that needs > to be checked. 1. client requests IP address 2. client receives IP address from server. DHCP issued DDNS update. 3. client hostname is changed using command: hostnamectl set-hostname "new hostname" On executing the renew command, nothing happens on the server side - i.e. there is no change to the leases file. On executing release and renew commands, all works as expected. The client receives the same IP and DHCP server sends DDNS using new hostname of the client. ---- Outstanding questions (answers being sought): 1. When the client (with the new hostname) renewed, did it send the new hostname with the renewal? 2. When the client (with the new hostname) released and renewed, did the DHCP server remove the old hostname? It would also be useful to see a network packet trace of these two tests (one that includes the client-server dialogue and all DDNS updates).
On Thu Nov 23 14:07:49 2017, fdupont wrote: > I reviewed the code: > - I pushed a few cosmetics stuff, please pull and review them > Looks good, thanks. > - I have a question which IMHO requires an answer: > when T1 and T2 are not zero or derived from lifetimes > I think they should be reduced? Note this is really specific > to DHCPv6 as for DHCPv4 this is only one timer doing > everything. > > My concern is the the threshold is a bit high after a reuse > the returned T1/T2 can (mis)lead too far. Good catch. I modified shorten_lifetimes() to update the reply minimum values which are used later on to calculate T1/T2. Seems to work just fine. Please re-review.
On Tue Nov 28 15:22:08 2017, tmark wrote: > Ready for re-review. => look OK.