CA realm (SERVERS.COM) instances demonstrating intermittent high latency/loss (209.192.x.x)

I'm pretty sick and tired of dying to things beyond my control: such as the servers spiking up randomly with high latency (especially when large Eater of Worlds packs spawn). This has been the case for the past 2 days where I've noticed very large intermittent spikes of in-game latency (500-2000ms+). Typically using the in-game metrics view, my Latency is 24-30ms on average. Hopefully you are monitoring metrics from your server instances, so you know when server side things are out of spec on your end.... right?

Speed test taken at time of recent lag related death (normal throughput, 1Gb/down, 35Mb/up)



Here's the map where latency spiked, and related bug report:

"

- Connecting to instance server at 209.192.171.148:6112
- Generating level 82 area "MapWorldsAtoll" with seed 795,354,257
- You have entered Atoll.

- 2022/12/26 16:47:21 - Your bug report reference number is 3,835,732,225.

Traceroute to IP in question (which typically, doesn't demonstrate anything of use here, but non-ggg people always seem insistent on asking for it):



Let traceroute run for over 8.5 minutes and as usual, servers.com is demonstrating some perceivable loss (se-sjc1-l15-r2.servers.com:173.237.36.87).



TL;DR please fix (cpu? packetloss?) issues on CA realm instances. Thanks.

None of your traces shows any packet loss. Traces of 30 seconds are useless, as they provide next to no reliable data, which is why the stickies ask for 300-500 second traces.

Speed tests are useless to determine anything useful.

None of this is a game bug.

"

Aragorn14 wrote:

None of your traces shows any packet loss. Traces of 30 seconds are useless, as they provide next to no reliable data, which is why the stickies ask for 300-500 second traces.

Speed tests are useless to determine anything useful.

None of this is a game bug.

First, I get how the internet works more than you could possibly imagine. I only provide these useless bits of data because you are one of the people constantly asking for it (EVEN THOUGH YOU ARE NOT A GGG EMPLOYEE).

These aren't "30 seconds of traces": because of how ICMP(echo) rate limiting works with many modern devices, typically low volume of these administrative packets will get discarded (looking like artificial loss to the originating client). To better demonstrate the visible (obviously I can't view tunnels or transit traffic) path that packets are currently taking, at this point in time, in one-direction I have the delay between packets at 3 seconds. So the second trace is just over 8.5 minutes of ping packets, not "30 seconds". In those 8.5 minutes I am seeing a constant 30% of packet loss at the hop just before GGG's instance servers (within the servers.com network).

The reason for the speedtest is to demonstrate that my connection's throughput doesn't have something else going on (such as upload throughput abysmally low, due to issues on my end). It's a minor item, but does have relevance as to helping to answer the question "is my customer premise equipment functioning properly".

Second, this is most certainly a problem on GGG's side whether it be with their instance servers or with their upstream hosting provider and/or network transit and does need attention. Please stop aggressively dismissing people who post real problems on these forums.

If I had more useful data from GGG's servers I would provide it; but I am not a GGG employee so here I am asking GGG to investigate and resolve.


Thanks.

I suggest running Wireshark as you play to see if it can give you a better idea of what's going on.

I've been experiencing random spikes of 5k latency that cause the game to freeze for a few seconds and come to find out, it was the GGG server requesting a packet that it already acknowledged that it received.

We'll be on Seq/88695 and then all of a sudden it'll request seq/81478, which will start the bullshit cycle of Multiple Dup/Ack/Retransmissions.

"

Xiuying wrote:

I suggest running Wireshark as you play to see if it can give you a better idea of what's going on.

I've been experiencing random spikes of 5k latency that cause the game to freeze for a few seconds and come to find out, it was the GGG server requesting a packet that it already acknowledged that it received.

We'll be on Seq/88695 and then all of a sudden it'll request seq/81478, which will start the bullshit cycle of Multiple Dup/Ack/Retransmissions.

I've run wireshark as well as some of the Windows powertooling to isolate from a high level issues like this in the past, and it's just exhausting. Unrelated example: used process monitor to demonstrate a bad hotfix last league that was causing multiple-seconds of area load times due to a broken change with how the client was loading textures. This issue wasn't totally unplayable if you were using SSD/NVMe but many people (who aren't using low latency high throughput storage) would time out on trying to enter a map.. oof.

I happen to have other ongoing network/systems monitoring infrastructure in place, so I am keenly aware of my overall network health/hardware/details. This is very much a problem being exhibited within the GGG scope of addressing, even it's a problem outside of their instance and with their upstream provider (servers.com). GGG would need to reach out for network health support there.

There have been short lived issues like this in the past but frankly, while this issue exists (which has observable over the past 2 days), it's making me not want to bother playing anymore. The XP loss is what makes this ultimately beyond frustrating, as it can absolutely eliminate hours of work.

"

TwentyFiveEX wrote:

Traceroute to IP in question (which typically, doesn't demonstrate anything of use here, but non-ggg people always seem insistent on asking for it):

https://www.pathofexile.com/forum/view-thread/3072628

So it's actually GGG that ask for it in these cases......

im pretty sure at this point that this league just brought in some sort of bug on their end that causes some interactions to lag the servers out. its normal for leaguestart to act up a bit but i dont know when the last time was where i saw so many people all over the world having huge lag issues with no obvious connection issue. hopefully this gets fixed because its getting unplayable for me in certain times aswell.

"

HanSoloDK wrote:

"

TwentyFiveEX wrote:

Traceroute to IP in question (which typically, doesn't demonstrate anything of use here, but non-ggg people always seem insistent on asking for it):

https://www.pathofexile.com/forum/view-thread/3072628

So it's actually GGG that ask for it in these cases......

Are you ggg's employee? If not,you should ask payment for defending them that hard lol

For what it's worth.. Tue, Dec-27: did not observe significant spikes in PoE client latency. Note: Also no longer farming maps with Eater of Worlds mobs.. switched to Searing Exarch. Unsure if this had -anything- to do with intermittent instance latency, but noting this for GGG devs.

Thanks.