Another thing to look for is QoS. By default when you telnet or ping from MX1 to MX4, then both MX1 and MX4 puts your icmp & telnet into Best Effort queue. You may want to put Telnet into a better than best effort class if you want a better response for your telnet session, and leave icmp as is.
http://www.juniper.net/documentation/en_US/junos13.3/topics/reference/general/hw-cos-default-re-queues-reference-cos-config-guide.html Huan > On 18 Oct 2014, at 5:49 pm, Huan Pham <[email protected]> wrote: > > Hi, > > If the passing through traffic does not experience any issue, then I do not > think it is related to "asymmetric" routing, or a congestion on a particular > link in the multi paths that exist between MX1 & MX4. > > It is more likely related to a high CPU utilisation on MX4, or a RE > protection (policing/ rate-limiting) mechanism on MX4. > > Huan, > >> On 17 Oct 2014, at 6:52 pm, R LAS <[email protected]> wrote: >> >> Strange issue. >> >> Let put it simple, my customer have a network like this: >> >> MX1 --- wan -- MX3 >> | | >> MX2 --- wan -- MX4 >> >> MX are MX480. >> >> Pings from MX1 to a vip address active on MX4 are 40% lost. >> >> A telnet from MX1 to a vip address active on MX4 takes 20 to 50 seconds to >> open the first time, the second is faster. >> I suspect asymetric routing but it's still not clear (I do not manage the >> customer net). >> >> I'm told that passing through traffic do not experience any kind of issue. >> >> I'm thinking to ddos protection but still not got the output from the >> customer... >> Any other idea about the issue ? >> >> Greetings >> _______________________________________________ >> juniper-nsp mailing list [email protected] >> https://puck.nether.net/mailman/listinfo/juniper-nsp _______________________________________________ juniper-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/juniper-nsp
