Site errors, sluggishness, and crashes

[sickmint79]

This is driving me nuts.


Colo guys are saying it's nothing on their end.. say it might be a bad NIC. (server has one NIC with 4 ports)

however, when I ping within the rack, to or from any of the 8 servers. There's zero packet loss.

If I ping outside of the rack, to the gateway or any place beyond that.. there's 25-35% packet loss.

There's a single network cable coming into the rack.. and yet none of the other servers have issues at all.

not sure what else to do.

so within rack, server to server pings go through a local switch on the rack and have 0 loss, tcg server pinging out hits immediate 25-35% loss first hop, but it's the only server in the rack to do so? that does sound pretty perplexing...

 

[Lord Tin Foilhat]

See if they will flush the cache on the switch.

 

[Jack]

so within rack, server to server pings go through a local switch on the rack and have 0 loss, tcg server pinging out hits immediate 25-35% loss first hop, but it's the only server in the rack to do so? that does sound pretty perplexing...

yes, within the rack, there are no dropped packets. Out side of the rack, the TCG server randomly drops packets. At about a 30% drop rate.

No other servers have this issue.

See if they will flush the cache on the switch.

My local switch is unmanaged, and it doesn't appear to be a routing issue, If I change IP's on the TCG server, the issue persists.



The server NIC is a 4 port daughter card.. we've tried other ports on the card, but have not tried a different NIC.

I'll ask if the COLO has a USB NIC I could use to test with.

 

[Mook]

Jack, I appreciate all that youre doing and the time youre putting in to fix this.

 

[Jack]

So it appears that I fixed it..

I'll wait for a bit, to make sure

 

[Mr_Roboto]

bad Ethernet cable?

 

[Jack]

Before and after, attached

 

[Jack]

bad Ethernet cable?

nope, that was already swapped

 

[FESTER665]

Definitely feels better right now... Good stuff [MENTION=288]Jack[/MENTION]

 

[sickmint79]

oooh do tell.

 

[Jack]

oooh do tell.

I changed the adapter Speed & Duplex setting from Auto to 100Mb Full Duplex

no freaking clue why that worked, or why it caused issues in the first place... I hadn't logged into the server for a long time prior to the issues surfacing.

and there were no driver updates.

 

[FirstWorldProblems]

straight up HERO

TY SIR!

 

[Lord Tin Foilhat]

I changed the adapter Speed & Duplex setting from Auto to 100Mb Full Duplex

no freaking clue why that worked, or why it caused issues in the first place... I hadn't logged into the server for a long time prior to the issues surfacing.

and there were no driver updates.

Whatever configured switch it eventually passes off to may have been the culprit

 

[sickmint79]

Whatever configured switch it eventually passes off to may have been the culprit

i'm not a network guy. could the auto be it continually renegotiating for some reason then and dropping packets?

 

[Lord Tin Foilhat]

i'm not a network guy. could the auto be it continually renegotiating for some reason then and dropping packets?

Possibly. Switches use the MAC so an ip change wouldn't help but forcing the comm type on the client NIC would definitely take the auto negotiate out if the equation which seems to have fixed the issue.

 

[Jack]

Possibly. Switches use the MAC so an ip change wouldn't help but forcing the comm type on the client NIC would definitely take the auto negotiate out if the equation which seems to have fixed the issue.

Yeah, but we did swap ports on the NIC, which should have unique MAC addresses.. and that didn't help

 

[Mook]

Browsing from tapa right now and it's flying.

 

[Flyn]

The frizzers are now running through the jinglesnappers.

Yeah, yeah. Important thing is the site is operating MUCH better. Thanks Jack and all the guys who had input.

 

[Lord Tin Foilhat]

Yeah, but we did swap ports on the NIC, which should have unique MAC addresses.. and that didn't help

Maybe that NIC doesnt like the switch

 

[Jack]

Maybe that NIC doesnt like the switch

But it liked it for 5+yrs
why the hate?

 

[EmersonHart13]

It must be a woman?

 

[Lord Tin Foilhat]

Relationships dont always get stronger with time

 

[Thirdgen89GTA]

Fuck, I should have asked about the Speed & Duplex settings yesterday when I finally read this thread.

I ran into the same problem several years ago.

Switched was hard coded to 100/half. Server NIC was set to auto, except it mistakenly kept picking 100/full.

Cause all sorts of collisions. Though in our case the managed Cisco switch would shut the interface down due to excessive errors until we manually reset it that port.

We had a manual setting because the switch was a Copper switch, using a Copper2Fiber converter on both ends because it was a 1000ft run. Why they didn't just spring for a real Fiber NIC I don't know. Its not like they were a Mom&Pop shop.

 

[Spivitz]

Yeah, I think Tapa is cranking faster than Mobile, maybe that's normal

 

[Mr_Roboto]

Duplex oddly makes sense.