Details
-
Task
-
Resolution: Won't Do
-
Major
-
None
-
7.1.0
-
1
-
KV 2022-Jan
Description
// The test makes assumptions on the memory state of the system at storing
|
// items. In particular, we load some items until we hit the HWM for
|
// preparing for HT ejection.
|
// Now that we introduce a limit on checkpoint mem-usage (MB-47386), when
|
// loading data we need to simulate checkpoint mem-recovery, or we would
|
// never reach the HWM otherwise as we'll hit the checkpoint quota first.
|
// All fine so far, under couchstore all behaves as expected.
|
// Problem under magma is that it allocates a good chunk of the bucket quota
|
// when we persist for allowing chekpoint-mem-recovery. We hit the HWM by
|
// magma allocations that seem to be intermittently freed (maybe in magma
|
// threads?) as soon as we return from the load-data function. Thus pushing
|
// the system back below the HWM and invalidating the test pre-requirements.
|
// At the time of writing magma is also facing some major internal memory
|
// usage and tracking issues.
|
// For all that, I'm disabling the test for magma for now.
|
// @todo: fix and re-enable, tracked in MB-47481
|
if (isMagma()) {
|
GTEST_SKIP();
|
}
|