From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:47441) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1eXXYg-0001sc-IP for qemu-devel@nongnu.org; Fri, 05 Jan 2018 14:21:39 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1eXXYd-0007wQ-5I for qemu-devel@nongnu.org; Fri, 05 Jan 2018 14:21:38 -0500 Received: from mx1.redhat.com ([209.132.183.28]:60422) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1eXXYb-0007vG-Uc for qemu-devel@nongnu.org; Fri, 05 Jan 2018 14:21:34 -0500 Received: from smtp.corp.redhat.com (int-mx05.intmail.prod.int.phx2.redhat.com [10.5.11.15]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id DB67C2D1D3 for ; Fri, 5 Jan 2018 19:21:32 +0000 (UTC) From: Juan Quintela In-Reply-To: <20171208114241.GJ2403@work-vm> (David Alan Gilbert's message of "Fri, 8 Dec 2017 11:42:42 +0000") References: <20171201125813.1437-1-quintela@redhat.com> <20171201125813.1437-6-quintela@redhat.com> <20171208114241.GJ2403@work-vm> Reply-To: quintela@redhat.com Date: Fri, 05 Jan 2018 20:21:29 +0100 Message-ID: <87r2r4nmbq.fsf@secure.laptop> MIME-Version: 1.0 Content-Type: text/plain Subject: Re: [Qemu-devel] [PATCH v3 5/6] tests: Add migration xbzrle test List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: "Dr. David Alan Gilbert" Cc: qemu-devel@nongnu.org, lvivier@redhat.com, peterx@redhat.com "Dr. David Alan Gilbert" wrote: > * Juan Quintela (quintela@redhat.com) wrote: >> Signed-off-by: Juan Quintela >> --- >> tests/migration-test.c | 68 ++++++++++++++++++++++++++++++++++++++++++++++++++ >> 1 file changed, 68 insertions(+) >> >> diff --git a/tests/migration-test.c b/tests/migration-test.c >> index 841c89e28a..41dee78a9a 100644 >> --- a/tests/migration-test.c >> +++ b/tests/migration-test.c >> @@ -410,6 +410,20 @@ static void deprecated_set_speed(QTestState *who, const char *value) >> migrate_check_parameter(who, "max-bandwidth", value); >> } >> >> +static void deprecated_set_cache_size(QTestState *who, const char *value) >> +{ >> + QDict *rsp; >> + gchar *cmd; >> + >> + cmd = g_strdup_printf("{ 'execute': 'migrate-set-cache-size'," >> + "'arguments': { 'value': %s } }", value); >> + rsp = qtest_qmp(who, cmd); >> + g_free(cmd); >> + g_assert(qdict_haskey(rsp, "return")); >> + QDECREF(rsp); >> + migrate_check_parameter(who, "xbzrle-cache-size", value); >> +} >> + >> static void migrate_set_parameter(QTestState *who, const char *parameter, >> const char *value) >> { >> @@ -424,6 +438,10 @@ static void migrate_set_parameter(QTestState *who, const char *parameter, >> deprecated_set_speed(who, "12345"); >> } >> >> + if (strcmp(parameter, "xbzrle-cache-size") == 0) { >> + deprecated_set_cache_size(who, "4096"); >> + } >> + >> cmd = g_strdup_printf("{ 'execute': 'migrate-set-parameters'," >> "'arguments': { '%s': %s } }", >> parameter, value); >> @@ -673,6 +691,55 @@ static void test_precopy_tcp(void) >> g_free(port); >> } >> >> +static void test_xbzrle(const char *uri) >> +{ >> + QTestState *from, *to; >> + >> + test_migrate_start(&from, &to, uri); >> + >> + /* We want to pick a speed slow enough that the test completes >> + * quickly, but that it doesn't complete precopy even on a slow >> + * machine, so also set the downtime. >> + */ >> + /* 100 ms */ >> + migrate_set_parameter(from, "downtime-limit", "1"); > > ? That's 1 ms not 100ms as the comment says? Changed comment. > >> + /* 1MB/s slow*/ >> + migrate_set_parameter(from, "max-bandwidth", "1000000000"); > > That's still 1GB/s isn't it? Changed also. > Don't you need to reduce that if xbzrle is actually working otherwise > it'll complete? It don't complete on my setup (dunno with a faster machine). >> + migrate_set_parameter(from, "xbzrle-cache-size", "33554432"); > > That's 32M - why? I need to check that the command is running, no clue about _which_ one is better here. > The test continually rewrites 100M of data; only changing one byte/page > - so actually the fun way to do the xbzrle test is to leave this as 32M > here..... My idea here is that we somke test all the transports/commands. Not that I fully test everything. See on my next series, It already takes around 15 seconds, just adding stuff will make things go longer really fast. BTW, is there a trivial way to convince qtest/getestr whatevre to run all the migration tests in parallel? Later, Juan. >> + migrate_set_capability(from, "xbzrle", "true"); >> + migrate_set_capability(to, "xbzrle", "true"); >> + /* Wait for the first serial output from the source */ >> + wait_for_serial("src_serial"); >> + >> + migrate(from, uri); >> + >> + wait_for_migration_pass(from); >> + >> + /* 300ms it should converge */ >> + migrate_set_parameter(from, "downtime-limit", "300"); > > but first increase the xvzrle cache to 120M and give it a couple of > seconds delay; you might even find it completes without the downtime > limit schange; but you should still do that after a few seconds. > You can check the stats as well, you should get some xbzrle cache hits > as long as you raised the cache size. > > Dave > >> + if (!got_stop) { >> + qtest_qmp_eventwait(from, "STOP"); >> + } >> + qtest_qmp_eventwait(to, "RESUME"); >> + >> + wait_for_serial("dest_serial"); >> + wait_for_migration_complete(from); >> + >> + test_migrate_end(from, to); >> +} >> + >> +static void test_xbzrle_unix(void) >> +{ >> + char *uri = g_strdup_printf("unix:%s/migsocket", tmpfs); >> + >> + test_xbzrle(uri); >> + g_free(uri); >> +} >> + >> + >> int main(int argc, char **argv) >> { >> char template[] = "/tmp/migration-test-XXXXXX"; >> @@ -695,6 +762,7 @@ int main(int argc, char **argv) >> qtest_add_func("/migration/precopy/unix", test_precopy_unix); >> qtest_add_func("/migration/precopy/tcp", test_precopy_tcp); >> qtest_add_func("/migration/postcopy/unix", test_postcopy); >> + qtest_add_func("/migration/xbzrle/unix", test_xbzrle_unix); >> >> ret = g_test_run(); >> >> -- >> 2.14.3 >> > -- > Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK