OpenSwan Drop Out Issue

Home » CentOS » OpenSwan Drop Out Issue
CentOS 8 Comments

Hello,

I’m cross posting this from the OpenSwan mailing list, in case someone here can help.

We have two sites connected via OpenSwan 2.6.32-9 on CentOS 5, sharing 6
/24 subnets each (so 12 in total).

The problem we’re having is completely randomly, be it in the middle of the day, or in the middle of the night (so I don’t believe it’s traffic related), certain (and sometimes all) routes will drop. They usually recover after a few minutes, but it’s still long enough for our monitoring to detect downtime.

The configuration we have on each device is:

conn site-a
keyingtries=0
keylife=1h
ikelifetime=8h
left=1.1.1.1
right=2.2.2.2

leftsubnets={x.x.x.x/24,x.x.x.x/24,x.x.x.x/24,x.x.x.x/24,x.x.x.x/24,x.x.x.x/24}

rightsubnets={x.x.x.x/24,x.x.x.x/24,x.x.x.x/24,x.x.x.x/24,x.x.x.x/24,x.x.x.x/24}
pfs=yes
auto=start
authby=secret
dpddelay0
dpdtimeout0
dpdaction=hold
phase2alg

8 thoughts on - OpenSwan Drop Out Issue

  • Thanks, I’ve updated the config with the following:

    keylife m
    ikelifetime=2h

    I’ll see how that goes.

    In the mean time, any other suggestions would be greatly appreciated.

  • Not off the top of my head, but if I were you, I’d enable debugging of
    “control” and “dpd”. See man ipsec.conf (/plutodebug) and man ipsec_pluto.

  • CentOS 5 is also a bit old os. Is it possible to use newer version? (like CentOS 7 or CentOS 6?)

    Eero

    2016-02-09 19:52 GMT+02:00 Gordon Messmer :

  • So lowering the keylife / ikelifetime didn’t solve the problem. I’ve enabled debugging and I’ll see what it says.

    Unfortunately we can’t (easily) upgrade CentOS, do you believe that would make a huge difference though? Are the newer versions of OpenSwan *that *much more reliable?

  • Well. CentOS 5 is really near of it’s end of life. There is not much updates to kernel or openswan. You should at least try latest openswan version.

    Your issue looks like a bit network problem.

  • As I said though, there’s no lost ICMP packets, even when the IPSec tunnel drops out.

    I do notice a lot of these errors in the secure log though, would this be any indication of a problem? (I’m grepping for this specific error, they’re not the only messages in there).

    Feb 11 14:18:10 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x01f90e1d) not found (maybe expired)
    Feb 11 14:18:14 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xb3681486) not found (maybe expired)
    Feb 11 14:18:14 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x6ad588f5) not found (maybe expired)
    Feb 11 14:19:07 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xe05ced4d) not found (maybe expired)
    Feb 11 14:19:08 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x7cd46e9e) not found (maybe expired)
    Feb 11 14:19:38 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x07164936) not found (maybe expired)
    Feb 11 14:19:55 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x9e68c142) not found (maybe expired)
    Feb 11 14:19:58 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xcbb10063) not found (maybe expired)
    Feb 11 14:20:16 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x7a160d48) not found (maybe expired)
    Feb 11 14:20:26 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x18a63776) not found (maybe expired)
    Feb 11 14:21:11 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x09eb87c4) not found (maybe expired)
    Feb 11 14:21:11 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xb2438c9b) not found (maybe expired)
    Feb 11 14:21:15 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x04236e6a) not found (maybe expired)
    Feb 11 14:21:52 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x456f7468) not found (maybe expired)
    Feb 11 14:21:57 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x8ee90acd) not found (maybe expired)
    Feb 11 14:22:04 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xc6676973) not found (maybe expired)
    Feb 11 14:22:04 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xc3b43142) not found (maybe expired)
    Feb 11 14:22:30 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x37111e62) not found (maybe expired)
    Feb 11 14:22:35 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xb6e63098) not found (maybe expired)
    Feb 11 14:23:24 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xbd94fd66) not found (maybe expired)
    Feb 11 14:24:05 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x36f47642) not found (maybe expired)
    Feb 11 14:24:18 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0xababea68) not found (maybe expired)
    Feb 11 14:24:33 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x9088954e) not found (maybe expired)
    Feb 11 14:24:46 site-a pluto[10450]: “site-b/1×1” #803: ignoring Delete SA
    payload: PROTO_IPSEC_ESP SA(0x5f1ba8d3) not found (maybe expired)

  • I think they indicate that both sides are restarting the tunnel, and that site-b is sending a “delete” command as it restarts the tunnel, while site-a has already removed the tunnel. But that doesn’t tell us anything about why they’re doing that. Control debugging from both sides *should* make that clear, but you’ll have to either make sense of the complete logs or share them.

LEAVE A COMMENT