Updates From Guru3D Folders

Discussion in 'Folding@Home - Join Team Guru3D !' started by Svein_Skogen, Aug 14, 2010.

  1. J_J_B

    J_J_B Member Guru

    Messages:
    159
    Likes Received:
    0
    GPU:
    EVGA: 2xSLI 580HC, 480HC
    Comcast DHCP renewal issued with dd-wrt router

    Folding on my system has been way down since Monday evening but things seem to be stable again now. Nothing to do with the PC. Rather, the services on my dd-wrt enabled router kept going down every hour. That included the OpenVPN service that facilitates the site-to-site VPN tunnel connecting the router on my office desk with the router at home. All traffic through the router on my office desk (what the PC is connected to) is directed through this tunnel and to the internet through my home router's gateway. I do this because it isn't permissable to have a personal PC on the corporate network; I have complete separation of my personal PC and LAN subnets from the office LAN the way I have things set up. I was also unable to remotely manage the home router - all this while away on a short business trip.

    The cause was unexpected and took a while to find this morning. It turns out that, when I power cycled everything at home Monday night to replace a failing UPS, the way that Comcast's cable modem assists with DHCP requests changed (must have received an update from Comcast upon being powered up). What is happening now is the lease period for DHCP assignments to the router's WAN port is limited to only 60 minutes. I don't know what it was before, but it was certainly greater than 24 hours and since my router is set to reboot itself at 2:00 AM this was never a problem before... the lease was renegotiated each morning during router startup, resulting in the lease period being re-upped to that >24-hour period.

    Per dd-wrt forums, the default firewall rule set blocks DHCP traffic that originates on the internet-side of the WAN port. Apparently it is normal for DHCP communications to be initiated by a DHCP server, but my custom firewall rules had only been allowing responses by the server to communications initiated by the router's DHCP client. Again, this wasn't a problem before when my router itself was initiating a renegotiation of the lease upon boot up each morning, but now some degree of server-originating traffic was required in order to negotiate the lease many times 23 additional times per day.

    So... I had to add a NAT rule and a set of Policy rules to the router's firewall to permit DHCP traffic originating from outside the router to penetrate the firewall and interract with the router's DHCP service. Before doing this, the lease would not get renewed until it ran down to zero minutes, at which point dd-wrt would temporarily shut down its interfaces, bring them back up, and then (successfully) initiate a full DHCP lease request. People inside the house would never know anything happened but the temporary take down of the interfaces caused many key services to error out and close. Even my custom firewall rules were invalidated and the default firewall rules put back into place... blocking my remote access to the router.

    With the corretions now applied, every 30 minutes the router receives a DHCP command from Comcast's servers to renegotiate and extend the WAN address lease - ultimately re-upping it to 60 minutes.

    <Addition>
    Man, I was totally bummed out a few minutes ago... the pump for my CPU/MB cooling loop was down around 4000 RPMs, from a norm of 4700 RPMs, and oscillating +- 200 RPM. I figured I had a failing pump and had already submitted a order for a replacement when I popped the right side cover off and fiddled around. I barely touched the pump's connector and the pump stopped. Connector was also warm. Turned out to be a simple loose connection. Tightened things up and RPMs are up over 4700 again. Order has been cancelled and I'm back in the waiting line for a 580 HC again.
     
    Last edited: Nov 20, 2010
  2. ariskar

    ariskar Master Guru

    Messages:
    206
    Likes Received:
    6
    GPU:
    Nvidia RTX 2080
    I was folding nicly over 20k ppd with some bigadv nice units, I left my PC on and went for excursion, though now I noticed it's not producing anymore... Anyway it will be back folding by tomorrow mates :)
     
    Last edited: Nov 20, 2010
  3. J_J_B

    J_J_B Member Guru

    Messages:
    159
    Likes Received:
    0
    GPU:
    EVGA: 2xSLI 580HC, 480HC
    Sorry to here about loss of production ariskar. Was it the result of a power loss? I'll be away from office for 4 days later this week so I'll be setting up a UPS there to ensure uptime of my own server.
     
  4. ariskar

    ariskar Master Guru

    Messages:
    206
    Likes Received:
    6
    GPU:
    Nvidia RTX 2080
    PC was on. It was actually the internet connection down for more than 10 minutes. I am on a PPPoE connection type and it stopped redialing so I couldn't get new WU's for the GPU3 client. Therefore my GTX480 was idle, while I got lucky to get a new bigadv WU on SMP before connection loss, which was and is still being processed!

    So back to task! :D
     

  5. J_J_B

    J_J_B Member Guru

    Messages:
    159
    Likes Received:
    0
    GPU:
    EVGA: 2xSLI 580HC, 480HC
    Sounds a little similar to what happened to me this week... folding bigadv's on SMP2 never quit. It was just my 480 that ran into problems due to the inopportune network outages.
     
  6. J_J_B

    J_J_B Member Guru

    Messages:
    159
    Likes Received:
    0
    GPU:
    EVGA: 2xSLI 580HC, 480HC
    Well... I thought I had my router debugged on Friday but problems continued on Saturday and into half the day on Sunday. It seems that, by default, the latest version of Firewall Builder is set to use utilities that are not included on current dd-wrt builds. When I was testing a for a solution to last week's problems on Friday I did not see this because those utilities are no longer missing once my router completely mounts it's USB hard drive and runs all scripts. The thing is, the startup scripts which do this work AND initially apply the firewall script, produced by Firewall Builder, are running in the context of the router's default (initial) environment. It doesn't matter if those missing utilities become available while the startup scripts are running - those scripts won't "see" the adjusted environment unless they are terminated and re-run. So the firewall script was immediately erroring out, denying incoming VPN connectivity from my office location and therefore continuing to cut internet service to my folding PC.

    As if that were not enough, my mail services (also running on the router) were terminating every 20 minutes whenever I did get the firewall rules applied. There was a spam message coming into one of my mailboxes that would cause ASSP (the Anti-Spam SMTP Proxy) to crash. I'd love to know what the underlying cause of the crash was but in the mean time I have added an ASSP rule to block all SMTP traffic from that source. ASSP is wonderfull, if a little slow, but I tend to have occaisional problems like this whereas I'm running on such a restrictive platform (SOHO router's don't have many resources). The underlying Perl 5.8.8 install on dd-wrt is also quite a hack and leads to problems with certain ASSP functions.

    Anyway... this morning was the first one in 6 days that did not find internet connectivity cut to my folding PC and mail services down. Folding operations have truly been back to normal since around noon yesterday. It will take a couple of days for my apparant 24-hour folding rate to get back up to where it was.
     
    Last edited: Nov 22, 2010
  7. ariskar

    ariskar Master Guru

    Messages:
    206
    Likes Received:
    6
    GPU:
    Nvidia RTX 2080
    480 GTX folding 16000~16500 @ 810 Mhz/ 1.063V/ 2000 Mhz Vram @ 72~ 74 degrees C :D
    Anything more would need a high VCore raise, while it is still on stock cooler/VRM/PCB.

    Result from 750Mhz to 810Mhz is about 800~1400ppd more.

    Anything more than 813Mhz@1.063V would cause EUE instability errors.
     
  8. ariskar

    ariskar Master Guru

    Messages:
    206
    Likes Received:
    6
    GPU:
    Nvidia RTX 2080
    I will set my F@H clients off for some period, as my girlfriend is working on Gromacs development team for her university degree project and will use my computer as a testing client for her custom modified/ optimized Gromacs molecular analysis core as her laptop will be the server. Domain of study is concentrated on Cuda/OpenCL acceleration which may affect GPU3/GPUv7 upcoming clients. That may mean that if something good comes out of the research, or a modded client, G3D users will be the first to get hands on it.

    For the time being, I wish you luck and consistency my fellow G3D folders!
     
  9. iancook221188

    iancook221188 Ancient Guru

    Messages:
    1,725
    Likes Received:
    11
    GPU:
    GTX 670 SLI / GTX 460 SLI
    that sound pretty cool good luck with that ariskar hop to see you back here soon
     
  10. J_J_B

    J_J_B Member Guru

    Messages:
    159
    Likes Received:
    0
    GPU:
    EVGA: 2xSLI 580HC, 480HC
    More problems with Comcast earlier this week. At first DNS name resolution began failing for clients on both my home and routed office networks. Then the ability to connect to my network from outside ceased. On Sunday night I gave up... shut everything down (router, cable modem, etc.) and powered things back up Monday morning, at which time a new DHCP address was assigned to my home router and everything began working again. The SMP2 client carried on because it already had work to do but my GPU3 client was inactive for about 10 hours.

    Words and civilized language cannot decribe how much I hate Comcast. Even if I were not attempting this advanced networking "stuff", their customer service just plain reeks. I'm going to need to do some tests on my brother-in-laws DirectTV (satellite) connection to determine if that ISP's network security will permit me to do what I need to do. If so, I'll probably be switching when I move into a new house in 2011.
     
    Last edited: Nov 30, 2010

  11. Ghost15

    Ghost15 Active Member

    Messages:
    77
    Likes Received:
    0
    GPU:
    EVGA 465 SC 1GB
    Ever since my first DNS Issue with my ISP, I switched to http://www.opendns.com/ Just configure your router (or PCs) to use their DNS instead. I've never had another DNS outage since I Switched.

    Might be worth a shot.
     
  12. J_J_B

    J_J_B Member Guru

    Messages:
    159
    Likes Received:
    0
    GPU:
    EVGA: 2xSLI 580HC, 480HC
    It goes beyond that though... I thought about doing the same Sunday evening but as I was troubleshooting I lost all connectivity, incoming and outgoing.

    I actually do use OpenDNS, but only for client requests on my routed (tunnelled) offsite office subnet. I created a NAT rule in the home router's firewall that redirects DNS reqests from those clients to OpenDNS for the sake of blocking clients from viewing most social network sites, adult sites, etc. This is to prevent employees who are testing the wi-fi connectivity and VPN profiles on their company computers (the original purpose of my setting up my private office subnet) from surfing such sites.

    <EDIT> Thanks for suggestion though... I do like OpenDNS!
     
    Last edited: Nov 30, 2010
  13. Ghost15

    Ghost15 Active Member

    Messages:
    77
    Likes Received:
    0
    GPU:
    EVGA 465 SC 1GB
    Well... My video card (280 GTX OC) from BFG died last night. Since BFG is gone no RMA for me. That takes my Hex i7 out of the game completely. Don't have the money to replace the card yet, so my PPD are going to drop hard untill I can get that back up and running. Hopefully just a couple weeks, as I need this PC for my work too!
     
  14. Mr.Bigtime

    Mr.Bigtime Ancient Guru

    Messages:
    20,794
    Likes Received:
    8
    GPU:
    ASUS 1080TI
    my pc is in trouble so i wont be able to FOLD for 1-2 weeks :..(
     
  15. iancook221188

    iancook221188 Ancient Guru

    Messages:
    1,725
    Likes Received:
    11
    GPU:
    GTX 670 SLI / GTX 460 SLI
    hey Ghost i though that Palit was doing a deal to help out people with BFG warranties/RMA i think they where just making new card cheaper if you had a old BFG card try making a thread in the 'Videocards - NVIDIA' section in guru3d and see what they can come up with to help out

    and Mr.B what happen to your rig:bawl:
     

  16. Mr.Bigtime

    Mr.Bigtime Ancient Guru

    Messages:
    20,794
    Likes Received:
    8
    GPU:
    ASUS 1080TI
    crashed by a faulty THERMALTAKE PC CASE :(
     
  17. PantherX

    PantherX Folding@Home Team Leader Staff Member

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    That is really sad news Ghost15 and Mr.Bigtime :( Hope you guys can fix the problem quickly.
     
  18. Ghost15

    Ghost15 Active Member

    Messages:
    77
    Likes Received:
    0
    GPU:
    EVGA 465 SC 1GB
    Thanks iancook, I'm going to look into that. Worth a shot and probobly a better idea than taking it apart and baking it. I think the memory is shot so I doubt the baking would fix it.
     
  19. J_J_B

    J_J_B Member Guru

    Messages:
    159
    Likes Received:
    0
    GPU:
    EVGA: 2xSLI 580HC, 480HC
    Also sorry to hear about the problems, Ghost15 and Mr. Bigtime.

    Ghost15: how are things going with EVGA? I assume you took the 980x out of service to RMA your X58 Classified on account of the erronious memory multiplier??

    The network problems I've had as of late pale in comparison. At the moment I'm monitoring the Laing DDC 3.25 pump on my GPU cooling loop, the speed of which has dropped from about 4500 to 4200 RPMs. Cooling efficiency of the loop has not decreased and the pump sounds ok, but I'm going to keep a close eye on this thing since neither of my cooling loops have any advanced failure detection mechanisms.
     
  20. Ghost15

    Ghost15 Active Member

    Messages:
    77
    Likes Received:
    0
    GPU:
    EVGA 465 SC 1GB
    No, the 980x has the 280 in it, no video card = cannot boot, so no folding even on the CPU.

    Apparently it is just being mis-read by the programs. As a test I set the multiplier to 2:16 , overclocking the memory by 200+ mhz, the memory instantly failed when I hit the windows boot logo. If the memory was really running at 2:10 despite my setting, then 2:16 would have 'worked'. I think thats pretty good proof it is running at 2:14 but being read as 2:10 for whatever reason.

    Thanks for all the support. I'm not sure how the new card I purchased will do with folding compared to my 280 but for 195 - 40 MIR I couldn't beat the price. It's a EVGA 465 w/ lifetime warrenty. Hope to have it by next weekend to install
     

Share This Page