|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
10/30/2014 We experienced a series of bounces today with some new system code that had passed an isolated test, but which failed in real world due to another part of the system that only ran on Cyber1, not the isolated test system. Previous version restored, system is back up and stable.
Details: The change that was put in place this morning was the first attempt to solve some framing / corruption issues in CIU mode due to the speed of the systems new CPU. These changes necessitated some lower level changes to internal networking, which affected several other "new" areas of the system, specifically a systems only TUTOR command that allows access the internet via a -socket- interface. This interface is used, for example, to post the current system status and user count to the www.cyber1.org main page (as well as usage stats). So with these crashes and dumps we can locate the other areas that need to be updated with the new low level internal networking changes. 9/22/2014 Recent outage log: 09/22/2014 10:15pm Cyber1 available. 09/22/2014 11:20am From Paul Koning - "Ok, tests on the new cyber1 machine confirm the fix is valid. I started a dtcyber about 15 minutes ago (not running PLATO, just plain NOS) and it is still right on the correct time. I'll leave it for a while longer. Then I'll work on getting the auto-restart machinery set up again. Then I'll deal with having it start PLATO again tonight at 10:15 pm, which is how far ahead the broken clock counted before I took the system down. Meanwhile, the "blackbox" program is announcing that, so people know the system will be coming back tonight." 09/22/2014 10:58am From Paul Koning - "... it is actually an issue in the NOS monitor code (in MTR, the PP monitor which does the timekeeping). It has assumptions on how fast a PPU runs, and in our emulated world those are not valid. The code doesn't actually make any sense anyway, but the real problem is that, if the emulation is fast enough, it will at times count milliseconds several times. On my Mac laptop, it does that perhaps 10% of the time so time is a bit fast. On the new machine, it's SO fast that it happens all the time, and possibly not just one extra tick per tick but more than one. It's obvious if I turn on very detailed logging of the clock machinery: it shows consecutive reads of the clock channel often just one or two microseconds apart. In a real Cyber, that's completely impossible because the read alone takes 2 microseconds, never mind all the instructions around it. I created a new file of all the NOS fixes in one place (the one for correctly handling February in leap years if the last restart was in January, remember that? as well as this one). Initial test on my Mac looks good, I'm building a new NOS now and will copy that to the new machine when it's ready, then test there." 09/22/2014 8:00am Cyber1 has been taken down until the time synchronization issue is resolved. Minimum time down will be 15 hours as we have to allow the real world clock to catch up to where Cyber1/DTCyber last time-stamped, which was off by 14+hours. 09/21/2014 8:05pm We know the time is off by 4 hours. When that is fixed Cyber1 will have to be down for 4 hours for the clock to catchup. 09/21/2014 7:59pm From Mike Cochran - "Cyber1 should be available to the public at this point. I put a note in 'pbnotes' saying to expect some teething problems." Our new system has been christened 'Solaris', which is not a reflection of the host operating system running (which is actually Linux Fedora). 09/19/2014 8:00am This just in from Mike Cochran - "I may start on things tonight or tomorrow morning. The whole process takes about a full day. Shooting for having computer up and "paulworthy" by Saturday PM, which puts Cyber1 up Sunday at the earliest." Note, "paulworthy" means that the hardware will be in place with basics of OS up and disk restored, etc., and ready for DTCyber configuration and bootstrap. 09/19/2014 7:52am Some have misunderstood the situation. Here is the gist of an email sent to a user who felt that they needed to donate to alleviate the "current crisis". "Donations are welcomed, but there should be no misunderstanding, we are not in an emergency situation. We have a spare computer being prepped. We lack time at the moment, not resources. These website (and Facebook page) statements were that we appreciate donations from the past year that help us keep Cyber1 running and we will need to buy a new machine to be kept in the wings for when this new machine,currently being prepped, dies sometime in the future. There is no causal action to you donating today that would speed up or in any way help the situation right now." That said we still welcome and appreciate donations. 09/18/2014 9:37am The system is still down. We'd like to take this opportunity to thank all the people who have donated to Cyber1 over the last year. Your donations have helped keep Cyber1 powered up, and also defrayed the cost of purchasing the current replacement box to run our DTCyber emulator. Please consider donating if you have not already done so, we have to purchase another replacement box to keep a spare on hand (BTW, we do not accept offers of hardware, we like to start fresh and keep it secure). 09/17/2014 4:29pm Linux host computer "bullwinkle" mainboard failed around 5:30am Pacific time. System could take several days to bring back online due to spare computer needing to be prepped. Cyber1 staff and operator are volunteers and we all hold down jobs in the real world, which requires our dedicated attention 6am-6pm M-F. We appreciate your understanding and patience. 11/20/2010 5:29:00 PM DtCyber was running but not functional. I did a kill -6 but strangely enough did not get a corefile. -- TFD 2/19/2010 1:04:00 AM dtcyber hung (no response from dd60, dtoper couldn't do a shutdown). kill -6 on PID 16325... Restarted. System was down from 18:19 until a few minutes ago. 1/20/2010 8:14:00 AM Primary WAN for host network unstable due to weather related issues. Temporary adjustment to DNS to issue IP for secondary WAN which seems more stable. You should expect intermittent connectivity until further notice. 12/28/2009 11:15:00 PM System up. 12/28/2009 11:14:00 PM Crash Log: System wasn't responding again, dropped dtcyber and reloaded process ID 28868. 12/28/2009 8:17:00 PM System froze. CIU displays PNTB, NAM is blank. Waiting for word from console operator. 7/1/2008 Power is restored. Cyber1 is back online. 6/30/2008 5:00:00 PM Cyber1 could be down for a while. I drove home on my bike to see a fire truck in front of my house, which is NEVER a good sign. I was pleasantly relieved to find out it was the transformer on the pole behind my house and fortunately behind the creek, but the firemen didn't realize they couldn't get across the barrier, so they had to drive around. But the upshot of it is: there is a pole there that had a transformer on it. The transformer is gone, the pole is smoking, and I hear from the neighbors that after this one went, another one down the street went. So this could be... cyber1 could be down for a while. Power's out at my house. From my experience, this could take up to a day. 10/29/2006 The equipment is in its new Mountain View home and the main server, Monster, is up and running cyber1 successfully. Neither DSL line is functional yet, so cyber1 is still not available. 10/26/2006 Cyber1 is moving to Mountain View and will be off-line for a few days. The hope is to have it up and running again Nov 1 or earlier. 10/20/2006 Posted 4 homespun TrueType fonts, two created by Adam Baum, and two created by Paul Koning. They are available as downloads in the Installation section of the website. 7/31/2006 System is back up. PS was fine; it was the UPS. Monster has been running 24/7 for over 2 years, under non-server room (sometimes downright hot) conditions. To make conditions even tougher, the mainframe emulator runs cpu-bound, i.e. one cpu is always pegged at 100%. As controlfreak Gerard van der Grinten put it, "If you can't stand the heat, don't play dtcyber." In any case we have a good backup machine on hand. 7/31/2006 Awoke this morning to a quiet server. It turned off some time between midnight and now, and will not power on. Best guess is a bum power supply. Regardless of etiology, two things are clear: 1 This problem will not be fixed until I get off work this evening; 2 All likely scenarios result in having the system back up in some fashion tonight (e.g. borrowing a PS from the other machine while I obtain a replacement). So look to the system being back up around 9 pm tonight. 7/23/2006 Pterm 4.06 available for all. It adds delay support in ascii mode. We passed 900 registered users today. 7/4/2006 Pterm 4.01 available. It has ascii support, and can talk to Aaron's custom-built keyboards. 4/1/2006 Registration page has been working fine. Paul sent pterm 3.02 for mac, which adds page up and page down and a couple other minor keyboard fixes. 3/15/2006 Fixed 1/2 of a problem with the registration page; now people can at least register. Also added pterm for windows 3.02. 3/14/2006 Today we moved the webserver back to its home in the rack. While the physical system changed location, cyber1.org had been hosted elsewhere (thanks to Gilbert Coville for hosting the website while cyber1 was moving). Cyberserv. org is now running fine under two new static ip's, 70.232.147.67 (SBC) and 69.17.54.41 (Speakeasy), and cyber1.org is at 70.232.147.68. 1/8/2006 Fixed address for Paul's cvs server. Now anyone can have the source to build the most current version of pterm. 1/2/2006 System was down for a good portion of Saturday and for a short time on Sunday due to very long power outtages here in Redwood City due to severe weather. 8/21/2005 System was unavailable for a couple hours last night due to boisterous group s activities while Mike was cooking. Some pictures are available here. 8/20/2005 Our production system is exactly one year old today. Party in Redwood City. 8/17/2005 Pterm v 2.37 for everything is available. Fixed problems with Gooch Box. 7/7/2005 Pterm v 2.34 for everything is available. Big fix to cut and paste, making it much easier for the user to have the system interact with the rest of their computer. 7/5/2005 Pterm v 2.32 for everything is available. It adds a nice help screen that shows the various keyboard combinations needed for PLATO. 5/21/2005 Pterm v 2.26 for mac is available as a disk image file. It fixes numerous keyboard problems with shifted keys, the numeric keypad, etc. 5/13/2005 Pterm v 2.24, which adds cut-and-paste support, available for unix, mac, and windows. 5/1/2005 Pterm v2.20 available. It adds Gooch Box support. 4/20/2005 Pterm v2.10 betas with ppt support available for windows and linux. 4/17/2005 Spring cleaning and numerous upgrades performed. 4/16/2005 Rather than work on taxes, Paul has continued refining the betas of pterm v2.03 for windows or linux, adding the ability to open multiple windows from the 'file' menu, the ability to turn on '1200 baud emulation' (fun), and fixes for a variety of keyboard issues. Bill Galcher and Paul Resch are helping work on the mac version, which should be ready soon. 4/13/2005 Pterm for Windows v2.01 beta and pterm 2.01 for linux et al are available. Paul fixed the F10 problem and removed -delay- support which was potentially causing problems in avatar. 4/10/2005 Pterm for Windows v2.00 beta is available for testing. It is windows-native, so just unzip and put on your desktop (no more command line!) It does not work well in avatar. Post bug reports in fixlist. 3/27/2005 Pterm 1.23 for windows, mac, and unix now available. Main change is easier support for characters like '@' while on the system. We passed 500 users this week. It is now normal to login at night to find yourself on with another 30-50 users. Added a new picture in 'screenshots' showing the final version of nos running on dtcyber, with our system, cyber1, loaded. 3/3/2005 Last week Paul finished the fix for the dropped characters by greatly increasing the size of the input buffer. We also upgraded from NOS 2.1 to the final version of NOS, 2.8.7. Up to 454 registered users. 2/16/2005 Lost data, had to restore from backup from 2/12 11 AM; any users created since then will have to be re-created. 434 users. 12/9/2004 Pterm 1.21 for mac, windows, and unix available. 12/8/2004 Some probs with one of the wans. If you are lagging tonight, try pterm 67.126.103.80 5004. 12/7/2004 Some probs with pterm 1.20. The unix version should be fixed, now v 1.21. Rolled back windows to 1.14 until it too is fixed. 12/4/2004 Pterm 1.20 for both unix and Windows available. 346 registered users. 11/29/2004 Pterm 1.16 for linux/mac available. 334 registered users. 11/18/2004 Pterm 1.14 for mac available. 316 registered users. 11/3/2004 Pterm 1.14 for Windows and 1.13 for unix (functionally the same) available. 263 registered users. 10/20/2004 Pterm 1.12 for Windows and Linux/OS X is out. Paul solved some timing issues caused by the server being so much faster than an original cyber. 218 users. 10/10/2004 Installed dual-wan router. Pretty experimental, but then again so is our opteron/linux platform. So far so good. About 180 users. 9/28/2004 Pterm v1.10 is out. Full portal keymappings plus arrow keys. 9/27/2004 All is well. 150 users. Pterm v1.04 now supports portal keymappings. 9/21/2004 Network has been OK now for 24 hours. 9/19/2004 Now having some network problems x ~3 days characterized by increasingly poor performance until the network goes down completely. ISP is working on it. Circuit was rebuilt tonight, maybe that will do the trick. 9/12/2004 The screen refresh problem appears greatly improved. Only complex screens take more than one drawing. The system is FAST now. Screen erase problem in pterm for windows is fixed; new file is pterm version 1.02. 9/9/2004 Up to 102 users. Current groups include cerl, minna, amesrad, wright, uofdel, pdev, zbrat, shire, progeny, cdcper, cfreaks, nova, fsu. System is pretty stable. Empire is happening. Soon will reinitialize avatar. Lots of stories being told. 9/4/2004 Paul added touchpanel support to the system at the cyber emulator level. He's still working on a more permanent fix to the dropped characters problem, although the system is working quite well now. Added version numbers to pterm, and posted the latest pterms for both windows and linux on the website. Both are UTD as of today and are equal in features. We're up to 89 registered users. 9/2/2004 The new website is up. A friend, Wil Voss, did it for fun. Thanks Wil! Paul may have fixed the dropped characters problem with a simple, tiny adjustment to the timing of the mainframe emulator. I made the adjustment tonight, we'll see how things go. The system seems faster to me, too. 8/31/2004 The pterm client for X is fixed (until we find the next bug that is). This version supports the touch screen. Our windows version is lagging behind with this latest enhancement. We have 66 registered users. 8/27/2004 All sorts of good stuff is happening. We have source code for pterm for unix, and it doesn't require a fonts install. That means OS X users have a client, too, although we're looking forward to a native mac client. Still having dropped characters on the system. Made announce/pad/empirenote. Ray Ozzie and Chuck Miller logged in. Made a signon for Brian Dear. 8/26/2004 A good portion of the erase abort screen problem is fixed. Some more signons given out to a test group of people. I hit "U" today and there were 7 users. Empire and avatar authors gave permissions to use their wonderful old games. A whole bunch of old PLATO people found out about the project in the process. 8/22/2004 Gave out the first signons to a small group of controlfreaks who can mess around on the system and generate errors. Probably still a month away from general users 8/20/2004 Paul got the portal key mappings added to both the X and the windows clients. The clients are now usable by adventurous general users. We still need to add features. The dual cpu production machine is online and the production incarnation of the system is working. 8/15/2004 Screen refresh problem remains, and the client isn't finished either. Accounts for the user groups cerl, minna, and nova made. What used to be entire systems are now groups on cyber1. Given the number of people who have already registered even though the system is not up, cerl may end up being broken into a couple constituent groups if there are some natural break points. Have yet to make group zbrat as we haven't had anyone claim that heritage yet. Paul sped the system up considerably. The dual cpu production machine is almost...almost... ready to be brought on line. Archived the existing notesfiles and cleaned them up for cyber1 use. 8/5/2004 Paul is still working on the screen refresh problem, and Steve is cranking on Paul's linux/unix pterm client, adding portal key mappings. Mike got email up and running very easily tonight using Jim Hickstein's Mirapoint imapserver—it has a simple, wonderful interface. Lots of fun and good feelings from emailing various old PLATO authors and telling them about cyber1. 7/31/2004 Paul has made big progress on the system, and it is running much more smoothly, although we are still having 'graphics' difficulties. The cgi backend for the registration form is now working. Screenshots added to main page. 7/25/2004 This progress page added to the website. Although the website has not been announced yet, the logs show that entities other than our little team are accessing it, which is obviously fine, and rather amazing given that it has been up for only two days. But someone at 64.8.171.6 has been trying to break into the webserver. It's pretty blatant hacking, trying accounts like root, guest, test, user, admin, etc. Whoever you are, why don't you just email us instead. We put up this page to help satisfy curiosity. 7/24/2004 System can now be accessed via our homespun terminal client, a vast UI improvement. Local access only. 7/21/2004 Cyber1 runs for the first time, from the console only. It is a very cool and strange feeling to bring up the ancient system. It feels a little like a time capsule. Navigating the system from the console is tedious. Much of the system is not working. Notes is not working. Response is moderately slow. |