After updating poky to "master:4517cdbfafdc7c5fa16bb5d09ed70c8e0da6a8c8", when building an image after a while I get: ERROR: Can NOT get PRAUTO, exception timed out ERROR: Function failed: package_get_auto_pr It has happened three times so far. I am using the following relevant options in my local.conf: BB_NUMBER_THREADS = "8" PARALLEL_MAKE = "-j 8" PRSERV_HOST = "localhost:0" Target machine is i586 Linux.
The time of each run before it gave the error: 23m53.394s 2m41.952s 18m43.564s I run bitbake with time nice -n19. I suspect it may be due to http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=bfab986ccd59636d0b0e79309f020e55481b0567.
The errors disappears if I revert bfab986ccd59636d0b0e79309f020e55481b0567.
If you swap timeout=5 to timeout=30 in that commit are things any better?
It would probably help but I have already built everything without the patch. I need a reliable method to reproduce it first.
Having timeout of 30 seems to fix it for me.
The performance of the PR service is drastically improved as of http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=d5e860be4c12a9d6655665af391ffd6f9351f8eb so can you see if you still have issues after this change?
I still have same error after http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=d5e860be4c12a9d6655665af391ffd6f9351f8eb with timeout of 5.
(In reply to comment #6) > The performance of the PR service is drastically improved as of > http://git.yoctoproject.org/cgit.cgi/poky/commit/ > ?id=d5e860be4c12a9d6655665af391ffd6f9351f8eb so can you see if you still > have issues after this change? It seems to be even worse now. After d5e860be4c12a9d6655665af391ffd6f9351f8eb using timeout of 30 fails with the following error (and more frequently): ERROR: Can NOT get PRAUTO from remote PR service ERROR: Function failed: package_get_auto_pr build/cache/prserve.log contains: ERROR: database table is locked: PRMAIN_nohist NOTE: PRServer: stopping...
I've pushed a change making the retry match for "table is locked" which it didn't previously. I've also increased the timeout to 60s since its causing people problems. If you could see if these two patches combined, I'd appreciate it, thanks!
I found further problem and went into this more deeply. There is now a series of patches on the bitbake mailing list which should improve the performance of the PR Server once and for all and resolve this problem, even with a 5s timeout.
It works after applying the two patches for 30 second timeout and SQLite retry fix.
I think the various bitbake changes to improve the PR server responsiveness have resolved this, ending with commit http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=c7b342903208cfb4cefaa0ec339c6ddfcd590acc
Tested with "timeout = 5" and there ware no issues. Tested on poky dora/master:faf8f8660fae26a14578738b642fdeb08bcb9d54