Bug#860627: The issue looks like not enough RAM
On 04/25/2017 05:56 PM, Lucas Nussbaum wrote: > On 25/04/17 at 16:23 +0200, Thomas Goirand wrote: >> Hi, >> >> As much as I can tell, it looks like to me that there's random crashes >> on i386 because of the lack of RAM. It looks like the test suite needs >> more than i386 can offer. Indeed, removing the --parallel option when >> running the tests seem to improve the situation. Though I'm not convince >> that removing that option is the thing to do. Anyway, I don't see a real >> user using a machine with 4GB as compute host these days. Even if >> someone would, they could well use the pre-built package, or not attempt >> to run tests when building. >> >> Therefore, I'll be lowering this bug severity to important, until I can >> get a reply from upstream. >> >> Your thoughts? > > Well the machine had 256 GB of RAM, but of course on i386 that's only > 4 GB per process. > > Lucas Yes, and OpenStack's gating is made with 8GB VMs. Anyway, I still stand by the fact that i386 has very little importance for something like OpenStack. I very much would prefer to invest time on ppcel64 and arm64. Cheers, Thomas Goirand (zigo)
Bug#860627: The issue looks like not enough RAM
On 25/04/17 at 16:23 +0200, Thomas Goirand wrote: > Hi, > > As much as I can tell, it looks like to me that there's random crashes > on i386 because of the lack of RAM. It looks like the test suite needs > more than i386 can offer. Indeed, removing the --parallel option when > running the tests seem to improve the situation. Though I'm not convince > that removing that option is the thing to do. Anyway, I don't see a real > user using a machine with 4GB as compute host these days. Even if > someone would, they could well use the pre-built package, or not attempt > to run tests when building. > > Therefore, I'll be lowering this bug severity to important, until I can > get a reply from upstream. > > Your thoughts? Well the machine had 256 GB of RAM, but of course on i386 that's only 4 GB per process. Lucas
Bug#860627: The issue looks like not enough RAM
Hi, As much as I can tell, it looks like to me that there's random crashes on i386 because of the lack of RAM. It looks like the test suite needs more than i386 can offer. Indeed, removing the --parallel option when running the tests seem to improve the situation. Though I'm not convince that removing that option is the thing to do. Anyway, I don't see a real user using a machine with 4GB as compute host these days. Even if someone would, they could well use the pre-built package, or not attempt to run tests when building. Therefore, I'll be lowering this bug severity to important, until I can get a reply from upstream. Your thoughts? Cheers, Thomas Goirand (zigo)