sometimes few flows get stuck in Hold state on fast tcp open/close traffic
Affects | Status | Importance | Assigned to | Milestone | ||
---|---|---|---|---|---|---|
Juniper Openstack | Status tracked in Trunk | |||||
Trunk |
Fix Committed
|
High
|
Praveen |
Bug Description
R3.0 Build 2709 Ubuntu 14.04 Kilo multinode
Have two VMs in a VN (20.1.1.3 and 20.1.1.4 in my case)
run hping3 from one to the other repeatedly
Ex :
hping3 -S -p 22 20.1.1.4 -s 10000 -c 1000 -i u1000
It is seen that sometimes , one or two flows get stuck in Hold state
gcore is in 10.204.
root@nodek1:~# cat flow.log
Flow table(size 68157440, entries 532480)
Entries: Created 17999 Added 17997 Processed 17997 Used Overflow entries 0
(Created Flows/CPU: 13898 2631 425 251 156 148 136 88 4 3 1 4 0 0 3 0 142 48 19 14 6 8 5 7 1 0 0 0 0 1 0 0)(oflows 0)
Action:F=Forward, D=Drop N=NAT(S=SNAT, D=DNAT, Ps=SPAT, Pd=DPAT, L=Link Local Port)
Other:
Flags:E=Evicted, Ec=Evict Candidate, N=New Flow, M=Modified
TCP(r=reverse)
Index Source:
-------
18104 20.1.1.3:10151 6 (3)
(K(nh):45, Action:H, Flags:, TCP:S, S(nh):0, Statistics:3/162 UdpSrcPort 0
214368<=>221348 20.1.1.3:22 6 (3->0)
(K(nh):45, Action:N(SD), Flags:, TCP:, S(nh):45, Statistics:
221348<=>214368 10.204.
(K(nh):5, Action:N(SD), Flags:, TCP:, S(nh):5, Statistics:
257104<=>479664 20.1.1.3:5031 1 (3)
(K(nh):45, Action:F, Flags:, S(nh):45, Statistics:
292080<=>305024 10.1.1.3:5031 1 (6)
(K(nh):20, Action:F, Flags:, S(nh):70, Statistics:
305024<=>292080 20.1.1.3:5031 1 (6)
(K(nh):20, Action:F, Flags:, S(nh):70, Statistics:
467336 20.1.1.3:10152 6 (3)
(K(nh):45, Action:H, Flags:, TCP:S, S(nh):0, Statistics:3/162 UdpSrcPort 0
479664<=>257104 10.1.1.3:5031 1 (3)
(K(nh):25, Action:F, Flags:, S(nh):25, Statistics:
root@nodek1:~#
Seeing the issue during Port-Tuple testing, whenever SI is added/removed from the policy.