From a06972a459de7cbd43cb683a12ba768f472863e2 Mon Sep 17 00:00:00 2001 From: Patrick Farrell Date: Fri, 15 Oct 2021 11:17:44 -0400 Subject: [PATCH] LU-15113 tests: Add margin to 398g MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit Every once in a great while, some other operation I can't identify triggers a single write RPC to a different file in test 398g on Gatekeeper testing. This has nothing to do with the test itself, but does cause it to fail occasionally.  An easy solution that isn't too bad for the test is to add a margin of +1 RPCs to account for this. Only modifies sanity, so trivial is OK. test-parameters: trivial Signed-off-by: Patrick Farrell Change-Id: I3936077cb60259653628ed26b01470ff529b0272 Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/45260 Reviewed-by: Andreas Dilger Reviewed-by: Vitaliy Kuznetsov Tested-by: jenkins Tested-by: Maloo --- lustre/tests/sanity.sh | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index f6ab132d..3fb9f3e 100755 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -27465,7 +27465,10 @@ test_398g() { # LU-13798 # We look at the "8 rpcs in flight" field, and verify A) it is present # and B) it includes all RPCs. This proves we had 8 RPCs in flight, # as expected for an 8M DIO to a file with 1M stripes. - [ $pct -eq 100 ] || error "we should see 8 RPCs in flight" + # NB: There is occasionally a mystery extra write RPC to a different + # file. I can't identify why that's happening, so we set up a margin + # of 1 RPC here, ie, 8/9 RPCs at this size, or ~88% + [ $pct -gt 87 ] || error "we should see 8 RPCs in flight" # Verify turning off parallel dio works as expected # Clear rpc stats -- 1.8.3.1