[FROG] ospf6d LSA's disappearing, probably stuck MaxAge LSA

[Leon Meßner]

Hi,
we recently moved two OSPFv3 routers from quagga 0.99.23 to frr-6.0.2
and are now having two (perhaps unrelated) issues. Issue one is that
link-states from one FRR (R1) do only appear very briefly in the
databases of other routers. This issue is the pressing one.
The other issue is that E1 routes from some routers stay in the ospf6d
route table although they have been flushed from the area already.

What i observe about issue one: Every LSRefreshTime, when restarting R1
or when "clear ipv6 ospf6 interface INTERFACE" is run, i see R1's LSA's
appear very briefly in neighbors databases. I also see routes related
to R1 appear in the ospf6d table.  After a few seconds these disappear
again. Looking at R1's "show ipv6 ospf6 database self-originated"
everything is there all the time.

What i see in debugging on some other Router (R2):

OSPF6: LSUpdate received on bond0.84
OSPF6:     src: fe80::3efd:feff:fea2:145a <-R1
OSPF6:     dst: ff02::5
OSPF6:     OSPFv3 Type:4 Len:108 Router-ID:192.109.82.2 <-R1
OSPF6:     Area-ID:0.0.0.0 Cksum:452b Instance-ID:0
OSPF6:     Number of LSA: 1    
OSPF6:     [Router Id:0.0.0.0 Adv:192.109.82.2]
OSPF6:     Age:    1 SeqNum: 0x800002b4 Cksum: 2e43 Len: 88 
OSPF6: ospf6_install_lsa Install LSA: [Router Id:0.0.0.0 Adv:192.109.82.2] age 1 seqnum 800002b4 in LSDB.
... flooding of that LSA ...
OSPF6: LSUpdate received on tun0
OSPF6:     src: fe80::f101:2   
OSPF6:     dst: fe80::f101:1   
OSPF6:     OSPFv3 Type:4 Len:108 Router-ID:192.109.82.24
OSPF6:     Area-ID:0.0.0.0 Cksum:7bab Instance-ID:0
OSPF6:     Number of LSA: 1    
OSPF6:     [Router Id:0.0.0.0 Adv:192.109.82.2]
OSPF6:     Age: 3600 SeqNum: 0x8000040c Cksum: 580c Len: 88 
OSPF6: ospf6_install_lsa Install LSA: [Router Id:0.0.0.0 Adv:192.109.82.2] age  268139 3600 seqnum 8000040c in LSDB.
OSPF6: LSA disappearing: [Router Id:0.0.0.0 Adv:192.109.82.2]
OSPF6: Schedule SPF Calculation for 0.0.0.0
OSPF6: SPF: Scheduled in 0 msec 
OSPF6: LSA disappearing: [Router Id:0.0.0.0 Adv:192.109.82.2]
... flooding of that LSA ...

So there is a MaxAge LSA that has a "higher" SequenceNumber.
Probably from a prior boot. Should FRR on R1 raise its SequenceNumber
for this LSA to 0x8000040d? Is this likely a problem on R1?

I also noticed, that LSAcks for MaxAge LSAs are multicast to either
ff02::6 or ff02::5 depending on DR status. Is this correct bahavior, or
should they be unicast?
I tried to reproduce issue one via FRRouting/topotests, but haven't
succeeded until now. Probably because i didn't manage to get a stuck
MaxAge LSA in my test instance.
Any help would be really appreciated. If more information is
needed, i'm happy to supply more logging and help with debugging.

Regards,
Leon