Hi,
the upgrade notes for Nautilus [0] contain this section:
Running nautilus OSDs will not bind to their v2 address
automatically. They must be restarted for that to happen.
Regards,
Eugen
[0] https://docs.ceph.com/en/latest/releases/nautilus/#instructions
Zitat von Mark Kirkwood <markkirkwood@xxxxxxxxxxxxxxxx>:
We have upgraded one of our Ceph clusters to Nautilus. We have run
into 2 issues that are causing osds to flap. I'll cover the 1st one
here, this one we solved but it raises an interesting question that
might bear on the 2nd one (will post that next).
After upgrading deb packages to Nautilus and restarting the mons and
mgrs we worked through restarting the osds. We started to see some
of them flap and saw this in the osd log (many times):
2024-07-31 11:03:33.264 7f22ab6e0700 0 --1-
[v2:[2404:130:8020:5::73]:6820/220732,v1:[2404:130:8020:5::73]:6821/220732]
>> v1:[2404:130:8020:5::103]:6909/2987374 conn(0x555a5bbbb180
0x555a4172c800 :-1 s=OPENED pgs=144 cs=3 l=0).fault initiating
reconnect
And later (usually a single line):
2404:130:8020:5::135]:6903/2933993 conn(0x555ab7b03200
0x555a099fa000 :-1 s=CONNECTING_SEND_CONNECT_MSG pgs=154 cs=4
l=0).handle_connect_reply_2 connect got BADAUTHORIZER
Examining the code showed:
markir@zmori:/download/ceph/src/ceph$ find . -type f -exec grep -l
"initiating reconnect" {} \;
./src/msg/simple/Pipe.cc
./src/msg/async/ProtocolV1.cc
./src/msg/async/ProtocolV2.cc
markir@zmori:/download/ceph/src/ceph$ vi src/msg/async/ProtocolV2.cc
markir@zmori:/download/ceph/src/ceph$ find . -type f -exec grep -l
"got BADAUTHORIZER" {} \;
./src/msg/simple/Pipe.cc
./src/msg/async/ProtocolV1.cc
Which led us to suspect that the osds were using the v1 msgr
protocol (ceph osd dump seems to validate this). We hoped that once
we enabled the v2 msgr that this error would vanish. This appears to
have happened.
So my question is this: looks like there is something wrong with
communications via v1 protocol post upgrade - is that expected?
Regards
Mark
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
