As an FYI, I'm sitting in a meeting now were we are pulling a check-in to use a re-entrant version of the function for our latest build to fix this issue. You should expect a fix in 11.2.4 unless something goes wrong.
Regards, Dan -----Original Message----- From: u2-users-boun...@listserver.u2ug.org [mailto:u2-users-boun...@listserver.u2ug.org] On Behalf Of John Hester Sent: Thursday, February 27, 2014 4:59 PM To: U2 Users List Subject: Re: [U2] Recommended 11.1.point release to upgrade to. We migrated from UV 10.2.7 on RedHat 5.1 x86 to 11.1.13 on RedHat 6.4 x64 last November. I've since run into a bug that can reveal itself when tty processes are terminated. It could be unique to linux, but you may want to watch for it. The symptoms are that terminated UV tty processes disappear from LISTU and PORT.STATUS, but continue to exist in the process table and consume a UV per-seat license. I discovered the issue when we ran out of licenses after a couple of months of uptime and I initially couldn't figure out where they went. Running the "ps" command at the OS level revealed their existence. I'm guessing the issue occurs on maybe 1 out of 200 or 300 tty sessions. The root cause of the issue is the localtime() function being called from the signal handler. The localtime() function is not POSIX async-signal-safe, which means it can't be safely called from there. The function acquires a lock which may already be held by the process that was interrupted by the signal if it too was in localtime(). When this happens, a deadlock is the result and the process is in limbo forever. I was able to easily work around the issue by having cron run a script after hours every day to clean up any hung UV processes and recover the licenses. I opened a ticket with Rocket, and they're planning to include a fix for the issue in 11.2.4. I'm happy to provide my workaround script to anyone who runs into this on a linux or unix box. Unfortunately, I'm not sure how one would craft a workaround on Windows. -John -----Original Message----- From: u2-users-boun...@listserver.u2ug.org [mailto:u2-users-boun...@listserver.u2ug.org] On Behalf Of Charles Stevenson Sent: Thursday, February 27, 2014 12:34 PM To: U2 Users List Subject: Re: [U2] Recommended 11.1.point release to upgrade to. Reporting back on a question I asked a few months ago. We finally upgraded from uv10.2.10 to uv11.1.15 mid-February. Delays were for internal business reasons having nothing to do with UV or the upgrade project itself. Platform is Windows 2003, which will be upgraded later this year. I'm pushing for Linux but not holding my breath. Primary goal was fear of falling off the back end of maintenance. I would have liked 11.2, but it is too new. There is no pressing business need to be an early adapter. Stability overrides. Because several years ago many here on both business & IT sides suffered a UV upgrade that caused the worst disaster I've ever seen a production system take. We restored from backup, losing 2 days of production data. That is the only problem I have ever seen with any UV upgrade. But I performed the one that went bad so I, personally, can't afford a 2nd with this same audience. So reluctance, nay, fear was high; regression testing, extensive; time between upgrades, long. (I won't get 11.2 for years unless we migrate to Linux.) Since every point release potentially introduces new bugs as well as fixes, I hesitated going to the latest 11.1.x, and toyed with going to a lesser one that more people are running on, pain-free. In the end we opted for the newest at that time, 11.1.13. If we ever have an issue, Rocket would probably put the fix in the next release & we'd have to install the cummulative changes, anyway. So we might as well test for as much as possible up front. Most regression testing was on 11.1.13. By the time we were ready to install, 11.1.15 was available. There did not appear to be much that affected us in -.14 & -.15, so I installed -.15 on the test system. Then mid-February I moved production from 10.2.10 to 11.1.15. Due to prior disaster, rollback-readiness to return to 10.2.10 was important. I exercised that a couple times on dev. Issues, comments: No issues during regression testing. The (default) uvhome is now c:\u2\uv instead of c:\ibm\uv. I chose to do a "new install" instead of upgrade. Permissions when installing 11.1.15 on production were tighter than when I installed 11.1.13 on dev. I don't know why. I like tight permissions, so I left them & it's ok. Had to be careful to allow update permissionw wher I created the new uv\errlog. MAKE.MAP.FILE had errors on both dev & prod after 11.1.15 install. I re-catalogued a couple subroutines it cared about and it seems to be ok. It wasn't a permissions problem. Gracious thanks to those on this list who offered advice, Chuck Stevenson On 9/25/2013 12:27 AM, Charles Stevenson wrote: > We're finally going to upgrade from 10.2.10 to 11.1.[something]. > > But which point release? > > We're on Win2003. (Linux next year. Baby steps.) . . . _______________________________________________ U2-Users mailing list U2-Users@listserver.u2ug.org http://listserver.u2ug.org/mailman/listinfo/u2-users _______________________________________________ U2-Users mailing list U2-Users@listserver.u2ug.org http://listserver.u2ug.org/mailman/listinfo/u2-users _______________________________________________ U2-Users mailing list U2-Users@listserver.u2ug.org http://listserver.u2ug.org/mailman/listinfo/u2-users
