These forums have been archived and are now read-only.

The new forums are live and can be found at https://forums.eveonline.com/

EVE General Discussion

 
  • Topic is locked indefinitely.
 

Regarding E-YJ8G last night

First post First post First post
Author
Previous Topic Next Topic
MeBiatch
GRR GOONS
#61 - 2013-11-23 18:02:58 UTC
James Amril-Kesh wrote:
Open up monitor during a fight like this, and what do I see?

The client is fetching bounties and killrights. Again and again and again.

Nobody ******* needs these in a large null battle. Killrights are completely irrelevant, and bounties can be payed out later/ Seriously, why are bounties still payed out instantly, when NPC bounties are on a 20 minute tick for the exact reason of avoiding excess server load?



good point i do wonder how much all those bounty calcs are slowing server load.

i would also support the payout per regular tick.

There are no stupid Questions... just stupid people... CCP Goliath wrote:

Ugh ti-di pooping makes me sad.
MeBiatch
GRR GOONS
#62 - 2013-11-23 18:07:42 UTC
what if instead of more poeple = tidi, more people = the opening of a temp black hole that kills every ship in system.

There are no stupid Questions... just stupid people... CCP Goliath wrote:

Ugh ti-di pooping makes me sad.
CCP Explorer
C C P
C C P Alliance
#63 - 2013-11-23 18:14:04 UTC
baltec1 wrote:
Did anyone even bother to inform CCP the fight was going to happen?
Yes, the node was reinforced.

Erlendur S. Thorsteinsson | Senior Development Director | EVE Online // CCP Games | @CCP_Explorer
Vera Algaert
Republic University
Minmatar Republic
#64 - 2013-11-23 18:16:23 UTC
What we need is a mechanic that causes, whenever too much mass is present on one grid, the entire grid collapse into a black hole.

.
CCP Explorer
C C P
C C P Alliance
#65 - 2013-11-23 18:22:01 UTC
MestariBation wrote:
CCP Veritas wrote:
Hiya folks, just a quick post about the node death that prematurely ended the fighting in E-YJ8G.

Technical investigation is ongoing, but it's clear that there was some run-away processing that took over 5 minutes of un-interrupted CPU time to complete. Our cluster is set up to assume that any node that is unresponsive for more than 3 minutes is dead and is removed forcefully from the cluster so that the show may go on elsewhere.

We're investigating why exactly that spike in execution happened and I'm hopeful we can knock it back in line. We'll be able to share more details of what exactly happened once a fix is deployed.
Just reimbuse everyhting and its fine.
Section 1.5 here http://community.eveonline.com/support/policies/eve-reimbursement-policy covers fleet fights.

Erlendur S. Thorsteinsson | Senior Development Director | EVE Online // CCP Games | @CCP_Explorer
CCP Explorer
C C P
C C P Alliance
#66 - 2013-11-23 18:23:58 UTC
BigSako wrote:
Laserak wrote:
CCP Veritas wrote:
It was processing a standard gameplay request from a player. This was not a case of developer interaction with the node.


Was it a Cheetah??


I've seen nodes crashed using a single Cloaky Tech 3 jumping through a gate. Not sure HOW they did that, but everytime you saw a T3 decloak, not moving and being unlockable, that would lock up the node.

That was already there in 2009 and I believe this bug is still out there.
A bug report with details would be super Big smile Please mention me and CCP Veritas in the report.

Erlendur S. Thorsteinsson | Senior Development Director | EVE Online // CCP Games | @CCP_Explorer
MeBiatch
GRR GOONS
#67 - 2013-11-23 18:41:57 UTC
CCP Explorer wrote:
baltec1 wrote:
Did anyone even bother to inform CCP the fight was going to happen?
Yes, the node was reinforced.


interesting sounds like a do while loop or something if one action was never fullfilled and caused the crash.

There are no stupid Questions... just stupid people... CCP Goliath wrote:

Ugh ti-di pooping makes me sad.
Kaj'Schak
Sebiestor Tribe
Minmatar Republic
#68 - 2013-11-23 19:25:56 UTC  |  Edited by: Kaj'Schak
CCP Explorer wrote:
MestariBation wrote:
CCP Veritas wrote:
Hiya folks, just a quick post about the node death that prematurely ended the fighting in E-YJ8G.

Technical investigation is ongoing, but it's clear that there was some run-away processing that took over 5 minutes of un-interrupted CPU time to complete. Our cluster is set up to assume that any node that is unresponsive for more than 3 minutes is dead and is removed forcefully from the cluster so that the show may go on elsewhere.

We're investigating why exactly that spike in execution happened and I'm hopeful we can knock it back in line. We'll be able to share more details of what exactly happened once a fix is deployed.
Just reimbuse everyhting and its fine.
Section 1.5 here http://community.eveonline.com/support/policies/eve-reimbursement-policy covers fleet fights.



let me get that right: your sevrer crashs, I land somewhere of battlefield on a safespot, not to have my fighterbombers returned as they are supposed to be with new features that were implemented some time ago when carriers/supercarriers crash? unable to reconnect?

ok
Razzor Death
The Scope
Gallente Federation
#69 - 2013-11-23 19:33:57 UTC
Kaj'Schak wrote:
CCP Explorer wrote:
MestariBation wrote:
CCP Veritas wrote:
Hiya folks, just a quick post about the node death that prematurely ended the fighting in E-YJ8G.

Technical investigation is ongoing, but it's clear that there was some run-away processing that took over 5 minutes of un-interrupted CPU time to complete. Our cluster is set up to assume that any node that is unresponsive for more than 3 minutes is dead and is removed forcefully from the cluster so that the show may go on elsewhere.

We're investigating why exactly that spike in execution happened and I'm hopeful we can knock it back in line. We'll be able to share more details of what exactly happened once a fix is deployed.
Just reimbuse everyhting and its fine.
Section 1.5 here http://community.eveonline.com/support/policies/eve-reimbursement-policy covers fleet fights.



let me get that right: your sevrer crashs, I land somewhere of battlefield on a safespot, not to have my fighterbombers returned as they are supposed to be with new features that were implemented some time ago when carriers/supercarriers crash? unable to reconnect?

ok


This has been the case for supers for a long time
BigSako
Aliastra
Gallente Federation
#70 - 2013-11-23 19:46:11 UTC
The difference is that we could have scooped our bombers, had we given a chance to logon were we crashed. In this case CCP messed up and put us in different spots, unable to recall bombers.
CCP Explorer
C C P
C C P Alliance
#71 - 2013-11-23 19:50:58 UTC
James Amril-Kesh wrote:
Open up monitor during a fight like this, and what do I see?

The client is fetching bounties and killrights. Again and again and again.

Nobody ******* needs these in a large null battle. Killrights are completely irrelevant, and bounties can be payed out later/ Seriously, why are bounties still payed out instantly, when NPC bounties are on a 20 minute tick for the exact reason of avoiding excess server load?
This information is actually not fetched from the location node, but rather a dedicated service that runs elsewhere in the cluster.

Erlendur S. Thorsteinsson | Senior Development Director | EVE Online // CCP Games | @CCP_Explorer
James Amril-Kesh
Viziam
Amarr Empire
#72 - 2013-11-23 19:52:42 UTC
CCP Explorer wrote:
James Amril-Kesh wrote:
Open up monitor during a fight like this, and what do I see?

The client is fetching bounties and killrights. Again and again and again.

Nobody ******* needs these in a large null battle. Killrights are completely irrelevant, and bounties can be payed out later/ Seriously, why are bounties still payed out instantly, when NPC bounties are on a 20 minute tick for the exact reason of avoiding excess server load?
This information is actually not fetched from the location node, but rather a dedicated service that runs elsewhere in the cluster.

And yet, when my game becomes unresponsive, this is almost always what's clogging up the backlog.

Enjoying the rain today? ;)
Leigh Akiga
Kuhri Innovations
#73 - 2013-11-23 19:52:46 UTC
There might be some big ones tonight if you CCP dudes are foreveralone at home like us capsuleers and wanted to do some realtime monitoring to see what all goes on Cool
Rattman
Deep Core Mining Inc.
Caldari State
#74 - 2013-11-23 20:01:52 UTC  |  Edited by: ISD Ezwal
CCP Veritas wrote:

Technical investigation is ongoing, but it's clear that there was some run-away processing that took over 5 minutes of un-interrupted CPU time to complete.


So take it all the capital pilots will get their billions of isk worth of bombers back they lost



Edit: fixed quote. ISD Ezwal.
Leigh Akiga
Kuhri Innovations
#75 - 2013-11-23 20:05:34 UTC
Rattman wrote:
So take it all the capital pilots will get their billions of isk worth of bombers back they lost


http://community.eveonline.com/support/policies/eve-reimbursement-policy

5. Any losses of any kind resulting from a large-scale player engagement are not covered by this reimbursement policy.

2. a. Only items destroyed along with the ship are eligible for reimbursement. Any items left in space (in wreck, containers, etc.) are not eligible for reimbursement.

lol
Icantspellwell
State Strategic Services
Templis Strategic Division
#76 - 2013-11-23 20:06:43 UTC
CCP Explorer wrote:
MestariBation wrote:
CCP Veritas wrote:
Hiya folks, just a quick post about the node death that prematurely ended the fighting in E-YJ8G.

Technical investigation is ongoing, but it's clear that there was some run-away processing that took over 5 minutes of un-interrupted CPU time to complete. Our cluster is set up to assume that any node that is unresponsive for more than 3 minutes is dead and is removed forcefully from the cluster so that the show may go on elsewhere.

We're investigating why exactly that spike in execution happened and I'm hopeful we can knock it back in line. We'll be able to share more details of what exactly happened once a fix is deployed.
Just reimbuse everyhting and its fine.
Section 1.5 here http://community.eveonline.com/support/policies/eve-reimbursement-policy covers fleet fights.


So let me get this straight CCP. Your server dies. Not only does it die but it dies so spectacularly that when it comes back and we log in we are flung to random points in the solar system unable to reconnect to our drones which should have automatically returned to our drones bays when we disconnected and you are not going to reimburse our fighterbombers that we would have never lost had your server worked properly? Instead of adding new features perhaps you should work on fixing your game so that we can get through one capital fight without your server dying a miserable death.
SurrenderMonkey
State Protectorate
Caldari State
#77 - 2013-11-23 20:20:42 UTC
Ishtanchuk Fazmarai wrote:
The solution is obvious.

People answer to incentives, but bad peformance is not incentive enough to stop piling every last possible ship on grid.

So, a incentive to stop piling ships on grid must be provided.

The war against poor performance in big battles is lost beforehand. No matter how much CCP raises the cap, every faction will try to outnumber the others and all combined will bring down the server.

So stop it altogether. Quality of experience is preferable to quantity of players bored to death as their game plays without them.

My suggestion: improve as much as possible the performance of 4,000 ships battles and add stacking penalties to fleet size and force proportion (that is, big fleets and fleets that massively outpower others will suffer increasingly harsh penalties).




While you're correct that the fight is lost before hand (owing largely to the quadratic nature of the scaling problem), that solution is garbage.

You were doing good when you said, "...an incentive to stop piling ships on grid must be provided..." but then you did an about face and instead tried to disincentivize that action instead of incentivizing another action.

A style of war that makes it strategically sensible to actually fight wars across a battle line (i.e., more than one system at a time), instead of at a battle point, would be far less offensive.

"Help, I'm bored with missions!"

http://swiftandbitter.com/eve/wtd/
Trillian Stargazer
Perkone
Caldari State
#78 - 2013-11-23 21:07:16 UTC
MeBiatch wrote:
Ting Mei wrote:
Veritas,

The problem is not only the node crash, but the TiDi system.

Even if we are all ok that is better to have TiDi than a systematic node crash as before, TiDi is really ruining the game little by little.

When you are in a fight, TIDI 10% is often activated, but the problem, is our brain not at 10%, and TIDI offer all the time for reinforcement to come, strategies to be discussed, etc ....

This really change the way of a fight ...

I'm very afraid CCP think TiDi is THE solution, and you are not working on something else ...

For me, don't know for others, even if on internal coms, most of players hate Tidi, that system is Ruining 0.0 fights little by little.


the only way to counter Tidi is to make deminishing returns for too much players or dps or something that makes it not worth it to bring 1000's of people to a fight... thats the problem is the blob...

i remember back in 06 when 200 people in a fight would cause black screens and node crashes now its 4000 people. the fact is they make it so you can play 4000 you guys will bring 5000 and so on... it really never ends...

there needs to be a mechanic for demishing returns.


There were only 1200 people in system last night + 3k drones and 1350 fighter bombers.

I have been in lowsec fights with more people and the node didnt die.
Gilbaron
The Scope
Gallente Federation
#79 - 2013-11-23 21:32:06 UTC
James Amril-Kesh wrote:
CCP Explorer wrote:
James Amril-Kesh wrote:
Open up monitor during a fight like this, and what do I see?

The client is fetching bounties and killrights. Again and again and again.

Nobody ******* needs these in a large null battle. Killrights are completely irrelevant, and bounties can be payed out later/ Seriously, why are bounties still payed out instantly, when NPC bounties are on a 20 minute tick for the exact reason of avoiding excess server load?
This information is actually not fetched from the location node, but rather a dedicated service that runs elsewhere in the cluster.

And yet, when my game becomes unresponsive, this is almost always what's clogging up the backlog.



guess you are confused between cause and effect here
Vincent Athena
Photosynth
#80 - 2013-11-23 21:35:24 UTC
CCP Explorer wrote:
James Amril-Kesh wrote:
Open up monitor during a fight like this, and what do I see?

The client is fetching bounties and killrights. Again and again and again.

Nobody ******* needs these in a large null battle. Killrights are completely irrelevant, and bounties can be payed out later/ Seriously, why are bounties still payed out instantly, when NPC bounties are on a 20 minute tick for the exact reason of avoiding excess server load?
This information is actually not fetched from the location node, but rather a dedicated service that runs elsewhere in the cluster.

It still has to be assembled into the packets sent to each client. If there are sufficient of them it uses up bandwidth.

Know a Frozen fan? Check this out

Frozen fanfiction