Node 3 seems to be used 50% more than node 1, but agreed that everything seems like it's getting traffic....
You still haven't identified what sort of monitor is in use, and if node 1 goes down/offline at any point...I'll give you a scenario all 3 getting normal load, but node 1 goes offline, either actually, or the lb things so, either way, it'll redirect all of 1's traffic over to 2/3....1 comes back online, but because of the round robin and the 3 min 'sticky' process, only every 3rd new user would be balanced over to node 1 and all existing users, as long as they use the system at least every 3 min's would stay on the rebalanced node they were put on when 1 went down.......that's a feasible scenario in my book but hard to prove out without logs, which you don't have. On Thu, Feb 1, 2018 at 10:17 AM, Thomas Miskiewicz <[email protected]> wrote: > They don’t have logging running because of the massive data load. But the > LB guy sent me the attached screenshot saying that it shows that all three > nodes get an equal amount of requests. > > If they really do, the I wonder why on one of the nodes I only see 2, 3 > users and and on the others an equal amount. Those 3 users are BTW AR Admin > and two technical users. Wired. Also speaking of amount of users I check > the license usage. Not aware of any other way checking. Are you? > > > On Feb 1, 2018, at 5:07 PM, LJ LongWing <[email protected]> wrote: > > I've always preferred to use least connections personally, but from what > I've read of the docs they recommend round robin....but round robin doesn't > explain how your LB would be putting everyone on one node...it should > literally go 1, 2, 3, 1, 2, 3 for new connections, and as you discussed, > keep you on your current server if you are within the timeout, but you > should have fairly even load across nodes....do they keep logs of the > monitor activity so you can review if the monitor is reporting a server as > down during specified periods? What monitor style are you using, ping, > tcpip port, something else? > > On Thu, Feb 1, 2018 at 9:00 AM, Thomas Miskiewicz <[email protected]> > wrote: > >> Hey LJ and thank you for the prompt reply. >> >> Talked to them already. We’re using round robin. I’m not an expert on LB >> but my understanding was that we would have about an equal amount of users >> on all AR Servers using it. >> >> The guy explained that once you connect there is a time out of 180 >> seconds. If you do something within the 180 seconds you will end up on the >> same server. If you don’t they will balance. Sounds plausible. Any yet >> what’s the likelihood that people then will be balanced on the second box >> and not the third. >> >> Does anyone there use round robin as well? How do your users get >> distributed? What are you Mid Tier settings, connection settings in >> particular? >> >> >> Thomas >> >> >> >> On Feb 1, 2018, at 4:49 PM, LJ LongWing <[email protected]> wrote: >> >> Thomas, >> You'll need to work with the LB team to identify what distribution method >> they are using....common options are 'round robin' in which it just simply >> points everything at each server in turn, 'least connections' where it >> tries to analyze how many are 'currently' connected to each node and send >> the traffic to the one with the least current load....you'll also want to >> check and verify that the monitor that you are using to determine if a node >> is online or not is functioning properly because if the LB monitor says a >> node is down, it obviously won't send any traffic to it...but if that >> monitor is faulty, it might be up and running but not reporting as online >> and can cause the scenario you described... >> >> On Thu, Feb 1, 2018 at 8:44 AM, Thomas Miskiewicz <[email protected]> >> wrote: >> >>> Hi Listers, >>> >>> we got a physical load balancer before the Mid-Tier servers and a >>> logical load balancer between the Mid Tier and our AR Servers. >>> >>> We noticed that on some days all the users land on only one AR Server. >>> On other days two. What’s the reason behind it? What do we need to >>> configure and to enforce equal distribution? >>> >>> >>> Thomas >>> -- >>> ARSList mailing list >>> [email protected] >>> https://mailman.rrr.se/cgi/listinfo/arslist >>> >> >> -- >> ARSList mailing list >> [email protected] >> https://mailman.rrr.se/cgi/listinfo/arslist >> >> >> >> -- >> ARSList mailing list >> [email protected] >> https://mailman.rrr.se/cgi/listinfo/arslist >> >> > -- > ARSList mailing list > [email protected] > https://mailman.rrr.se/cgi/listinfo/arslist > > > > -- > ARSList mailing list > [email protected] > https://mailman.rrr.se/cgi/listinfo/arslist > >
-- ARSList mailing list [email protected] https://mailman.rrr.se/cgi/listinfo/arslist
