From: Alexander Egorenkov Date: Fri, 22 Jan 2021 08:56:50 +0100 Subject: [LTP] [PATCH 1/1] swapping01: make test more robust In-Reply-To: References: <20210115143246.369676-1-egorenar@linux.ibm.com> <87zh14rrxm.fsf@oc8242746057.ibm.com> Message-ID: <87y2glxv3x.fsf@oc8242746057.ibm.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ltp@lists.linux.it Hi Li, Date: Fri, 22 Jan 2021 08:56:45 +0100 Message-ID: <87zh11xv42.fsf@oc8242746057.ibm.com> X-Trend-IP-HD: ip=[9.171.79.240]sender=(egorenar@linux.ibm.com)recipient= Li Wang writes: > Hi Alex, Petr, > > I'm going to apply Alex's patch after a round of testing[1]. > Because giving more time waiting for swap memory to > settle is no harm to test, at least help to get rid of failures > observed on LPAR so far. Great, thank you very much :) I also think it shouldn't make the test behave worse. > > But I still think the swapping01 not very precise to some degree. > We probably need to rewrite it someday but currently, I have no better > idea. 100% agree. It is very fragile. I was thinking maybe we could find out how much swap space is assigned to a *particular* process, maybe something in /proc/ can be used ? > > [1] Running a stress-ng to eat some available memory in the > background, then swapping01 easily hit the heavy-swap issue. > This could be proved that it influenced by deamon allocate/free > memory during testing. > > # stress-ng to eat memory after starting swapping01 > I also observed that with stress-ng. Regards Alex