Re: [ovs-discuss] Drops in OVS

2015-08-01 Thread Abhishek Verma
HI, Is my understanding below correct? On Fri, Jul 31, 2015 at 10:14 AM, Abhishek Verma wrote: > Hi Flavio, > > >> > Is this correct? >> >> No, there is a cache in the kernel which is periodically refreshed or >> empty at the initialization. When the first packet comes in the >> packet is que

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Abhishek Verma
Hi Flavio, > > Is this correct? > > No, there is a cache in the kernel which is periodically refreshed or > empty at the initialization. When the first packet comes in the > packet is queued to the userspace daemon which will update the > kernel's cache. Next packets matching the cached flow wi

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Flavio Leitner
On Fri, Jul 31, 2015 at 09:38:17AM +0530, Abhishek Verma wrote: > > > > > > > > I just see *4424 *packets getting lost. What are these packets? Any idea > > > when would we lose them? > > > > "lost" means that a queue from the kernel to userspace overflowed. It > > would ordinarily happen if there

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Ben Pfaff
On Fri, Jul 31, 2015 at 06:48:26AM +0530, Abhishek Verma wrote: > > OVS wouldn't normally drop packets. It would normally happen before or > > after OVS. > > > > OVS *can* drop packets at the kernel/user interface. Those would show > > up in "ovs-dpctl show" as "lost" packets. > > > > Are you loo

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Abhishek Verma
Ben, > OVS wouldn't normally drop packets. It would normally happen before or > after OVS. > > OVS *can* drop packets at the kernel/user interface. Those would show > up in "ovs-dpctl show" as "lost" packets. > > Are you looking at the OVS vport stats or the stats via e.g. "ifconfig"? > Both ar

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Ben Pfaff
On Fri, Jul 31, 2015 at 06:00:47AM +0530, Abhishek Verma wrote: > So how do i verify if the packet got dropped at the egress. In my case the > output is a single port. My counter increments, which means OVS will now > send the packet to the egress port. If the interface stats for the output > port

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Abhishek Verma
Great. So, its an ingress counter then. So how do i verify if the packet got dropped at the egress. In my case the output is a single port. My counter increments, which means OVS will now send the packet to the egress port. If the interface stats for the output port show zero drops then can i assu

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Ben Pfaff
On Fri, Jul 31, 2015 at 05:38:12AM +0530, Abhishek Verma wrote: > When i add a flow rule which matches on some source IP, dest IP and i give > a output port as an action, then if i see the packet counter incrementing > then can i be sure that the packet was successfully sent out by OVS? > > Is the

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Abhishek Verma
Thanks Flavio. I installed very specific flow-rules and checked the counter values. That way i checked the segment where there were losses. My question is this: When i add a flow rule which matches on some source IP, dest IP and i give a output port as an action, then if i see the packet counter

Re: [ovs-discuss] Drops in OVS

2015-07-30 Thread Flavio Leitner
On Thu, Jul 30, 2015 at 05:56:29AM +0530, Abhishek Verma wrote: > Hi, > > I have 3 switches connected via OVS like this: > > R1 -- R2 -- R3 > > I am doing an iperf test using TCP between R1 and R3. R1, R2 and R3 are > spread out geographically and i use a VXLAN tunnel to connect them. > > iperf

[ovs-discuss] Drops in OVS

2015-07-29 Thread Abhishek Verma
Hi, I have 3 switches connected via OVS like this: R1 -- R2 -- R3 I am doing an iperf test using TCP between R1 and R3. R1, R2 and R3 are spread out geographically and i use a VXLAN tunnel to connect them. iperf test shows me considerable packet losses. However, ovs-vsctl get interface stats is