[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-devel] [PATCH 4 of 6] tools: set migration constraints from cmdline
# HG changeset patch # User Olaf Hering <olaf@xxxxxxxxx> # Date 1363623457 -3600 # Node ID a897276ed24d434daf5a5d3d8543343a23df2f8a # Parent b3ed5c4b5078e5bf90d23f7b09eda727af0d7e74 tools: set migration constraints from cmdline Add new options to xm/xl migrate to control the process of migration. The intention is to optionally abort the migration if it takes too long to migrate a busy guest due to the high number of new dirty pages. Currently the guest is suspended to transfer the remaining dirty pages. This transfer can take a long time, which can confuse the guest if its suspended for too long. The new options allow to override the built-in default values, which are not changed by this patch. --max_iters <number> Number of iterations before final suspend (default: 30) --max_factor <factor> Max amount of memory to transfer before final suspend (default: 3*RAM) --abort_if_busy Abort migration instead of doing final suspend. The changes to libxl change the API, handle LIBXL_API_VERSION == 0x040200. TODO: eventually add also --min_remaining (default value 50) in a seperate patch v7: - remove short options - update description of --abort_if_busy in xl.1 - extend description of --abort_if_busy in xl help - add comment to libxl_domain_suspend declaration, props is optional v6: - update the LIBXL_API_VERSION handling for libxl_domain_suspend change it to an inline function if LIBXL_API_VERSION is defined to 4.2.0 - rename libxl_save_properties to libxl_domain_suspend_properties - rename ->xlflags to ->flags within that struct v5: - adjust libxl_domain_suspend prototype, move flags, max_iters, max_factor into a new, optional struct libxl_save_properties - rename XCFLAGS_DOMSAVE_NOSUSPEND to XCFLAGS_DOMSAVE_ABORT_IF_BUSY - rename LIBXL_SUSPEND_NO_FINAL_SUSPEND to LIBXL_SUSPEND_ABORT_IF_BUSY - rename variables no_suspend to abort_if_busy - rename option -N/--no_suspend to -A/--abort_if_busy - update xl.1, extend description of -A option v4: - update default for no_suspend from None to 0 in XendCheckpoint.py:save - update logoutput in setMigrateConstraints - change xm migrate defaults from None to 0 - add new options to xl.1 - fix syntax error in XendDomain.py:domain_migrate_constraints_set - fix xm migrate -N option name to match xl migrate v3: - move logic errors in libxl__domain_suspend and fixed help text in cmd_table to separate patches - fix syntax error in XendCheckpoint.py - really pass max_iters and max_factor in libxl__xc_domain_save - make libxl_domain_suspend_0x040200 declaration globally visible - bump libxenlight.so SONAME from 2.0 to 2.1 due to changed libxl_domain_suspend v2: - use LIBXL_API_VERSION and define libxl_domain_suspend_0x040200 - fix logic error in min_reached check in xc_domain_save - add longopts - update --help text - correct description of migrate --help text Signed-off-by: Olaf Hering <olaf@xxxxxxxxx> diff -r b3ed5c4b5078 -r a897276ed24d docs/man/xl.pod.1 --- a/docs/man/xl.pod.1 +++ b/docs/man/xl.pod.1 @@ -391,6 +391,21 @@ Send <config> instead of config file fro Print huge (!) amount of debug during the migration process. +=item B<--max_iters> I<number> + +Number of iterations before final suspend (default: 30) + +=item B<--max_factor> I<factor> + +Max amount of memory to transfer before final suspend (default: 3*RAM) + +=item B<--abort_if_busy> + +Abort migration instead of doing final suspend/transfer/resume if the +guest has still dirty pages after the number of iterations and/or the +amount of RAM transferred. This avoids long periods of time where the +guest is suspended. + =back =item B<remus> [I<OPTIONS>] I<domain-id> I<host> diff -r b3ed5c4b5078 -r a897276ed24d tools/libxc/xc_domain_save.c --- a/tools/libxc/xc_domain_save.c +++ b/tools/libxc/xc_domain_save.c @@ -804,6 +804,7 @@ int xc_domain_save(xc_interface *xch, in int rc = 1, frc, i, j, last_iter = 0, iter = 0; int live = (flags & XCFLAGS_LIVE); int debug = (flags & XCFLAGS_DEBUG); + int abort_if_busy = (flags & XCFLAGS_DOMSAVE_ABORT_IF_BUSY); int superpages = !!hvm; int race = 0, sent_last_iter, skip_this_iter = 0; unsigned int sent_this_iter = 0; @@ -1527,10 +1528,20 @@ int xc_domain_save(xc_interface *xch, in if ( live ) { + int min_reached = sent_this_iter + skip_this_iter < 50; if ( (iter >= max_iters) || - (sent_this_iter+skip_this_iter < 50) || + min_reached || (total_sent > dinfo->p2m_size*max_factor) ) { + if ( !min_reached && abort_if_busy ) + { + ERROR("Live migration aborted, as requested. (guest too busy?)" + " total_sent %lu iter %d, max_iters %u max_factor %u", + total_sent, iter, max_iters, max_factor); + rc = 1; + goto out; + } + DPRINTF("Start last iteration\n"); last_iter = 1; diff -r b3ed5c4b5078 -r a897276ed24d tools/libxc/xenguest.h --- a/tools/libxc/xenguest.h +++ b/tools/libxc/xenguest.h @@ -28,6 +28,7 @@ #define XCFLAGS_HVM (1 << 2) #define XCFLAGS_STDVGA (1 << 3) #define XCFLAGS_CHECKPOINT_COMPRESS (1 << 4) +#define XCFLAGS_DOMSAVE_ABORT_IF_BUSY (1 << 5) #define X86_64_B_SIZE 64 #define X86_32_B_SIZE 32 diff -r b3ed5c4b5078 -r a897276ed24d tools/libxl/Makefile --- a/tools/libxl/Makefile +++ b/tools/libxl/Makefile @@ -5,7 +5,7 @@ XEN_ROOT = $(CURDIR)/../.. include $(XEN_ROOT)/tools/Rules.mk -MAJOR = 2.0 +MAJOR = 2.1 MINOR = 0 XLUMAJOR = 1.0 diff -r b3ed5c4b5078 -r a897276ed24d tools/libxl/libxl.c --- a/tools/libxl/libxl.c +++ b/tools/libxl/libxl.c @@ -756,7 +756,8 @@ static void domain_suspend_cb(libxl__egc } -int libxl_domain_suspend(libxl_ctx *ctx, uint32_t domid, int fd, int flags, +int libxl_domain_suspend(libxl_ctx *ctx, uint32_t domid, int fd, + const libxl_domain_suspend_properties *props, const libxl_asyncop_how *ao_how) { AO_CREATE(ctx, domid, ao_how); @@ -777,8 +778,13 @@ int libxl_domain_suspend(libxl_ctx *ctx, dss->domid = domid; dss->fd = fd; dss->type = type; - dss->live = flags & LIBXL_SUSPEND_LIVE; - dss->debug = flags & LIBXL_SUSPEND_DEBUG; + if (props) { + dss->live = props->flags & LIBXL_SUSPEND_LIVE; + dss->debug = props->flags & LIBXL_SUSPEND_DEBUG; + dss->max_iters = props->max_iters; + dss->max_factor = props->max_factor; + dss->xlflags = props->flags; + } libxl__domain_suspend(egc, dss); return AO_INPROGRESS; diff -r b3ed5c4b5078 -r a897276ed24d tools/libxl/libxl.h --- a/tools/libxl/libxl.h +++ b/tools/libxl/libxl.h @@ -507,12 +507,31 @@ int libxl_domain_create_restore(libxl_ct void libxl_domain_config_init(libxl_domain_config *d_config); void libxl_domain_config_dispose(libxl_domain_config *d_config); +typedef struct { + int flags; /* LIBXL_SUSPEND_* */ + int max_iters; + int max_factor; +} libxl_domain_suspend_properties; + int libxl_domain_suspend(libxl_ctx *ctx, uint32_t domid, int fd, - int flags, /* LIBXL_SUSPEND_* */ + const libxl_domain_suspend_properties *props, /* optional */ const libxl_asyncop_how *ao_how) LIBXL_EXTERNAL_CALLERS_ONLY; +#ifdef LIBXL_API_VERSION +#if LIBXL_API_VERSION == 0x040200 +static inline int libxl_domain_suspend_0x040200(libxl_ctx *ctx, uint32_t domid, int fd, + int flags, const libxl_asyncop_how *ao_how) +{ + libxl_domain_suspend_properties props = { .flags = flags }; + return libxl_domain_suspend(ctx, domid, fd, &props, ao_how); +} +#define libxl_domain_suspend libxl_domain_suspend_0x040200 +#endif +#endif + #define LIBXL_SUSPEND_DEBUG 1 #define LIBXL_SUSPEND_LIVE 2 +#define LIBXL_SUSPEND_ABORT_IF_BUSY 4 /* @param suspend_cancel [from xenctrl.h:xc_domain_resume( @param fast )] * If this parameter is true, use co-operative resume. The guest diff -r b3ed5c4b5078 -r a897276ed24d tools/libxl/libxl_dom.c --- a/tools/libxl/libxl_dom.c +++ b/tools/libxl/libxl_dom.c @@ -1297,6 +1297,7 @@ void libxl__domain_suspend(libxl__egc *e dss->xcflags = (live ? XCFLAGS_LIVE : 0) | (debug ? XCFLAGS_DEBUG : 0) + | (dss->xlflags & LIBXL_SUSPEND_ABORT_IF_BUSY ? XCFLAGS_DOMSAVE_ABORT_IF_BUSY : 0) | (dss->hvm ? XCFLAGS_HVM : 0); dss->suspend_eventchn = -1; diff -r b3ed5c4b5078 -r a897276ed24d tools/libxl/libxl_internal.h --- a/tools/libxl/libxl_internal.h +++ b/tools/libxl/libxl_internal.h @@ -2249,6 +2249,9 @@ struct libxl__domain_suspend_state { xc_evtchn *xce; /* event channel handle */ int suspend_eventchn; int hvm; + int max_iters; + int max_factor; + int xlflags; int xcflags; int guest_responded; const char *dm_savefile; diff -r b3ed5c4b5078 -r a897276ed24d tools/libxl/libxl_save_callout.c --- a/tools/libxl/libxl_save_callout.c +++ b/tools/libxl/libxl_save_callout.c @@ -108,8 +108,8 @@ void libxl__xc_domain_save(libxl__egc *e } const unsigned long argnums[] = { - dss->domid, 0, 0, dss->xcflags, dss->hvm, vm_generationid_addr, - toolstack_data_fd, toolstack_data_len, + dss->domid, dss->max_iters, dss->max_factor, dss->xcflags, dss->hvm, + vm_generationid_addr, toolstack_data_fd, toolstack_data_len, cbflags, }; diff -r b3ed5c4b5078 -r a897276ed24d tools/libxl/xl_cmdimpl.c --- a/tools/libxl/xl_cmdimpl.c +++ b/tools/libxl/xl_cmdimpl.c @@ -3201,7 +3201,7 @@ static int save_domain(uint32_t domid, c save_domain_core_writeconfig(fd, filename, config_data, config_len); - int rc = libxl_domain_suspend(ctx, domid, fd, 0, NULL); + int rc = libxl_domain_suspend(ctx, domid, fd, NULL, NULL); close(fd); if (rc < 0) @@ -3361,6 +3361,7 @@ static void migrate_do_preamble(int send } static void migrate_domain(uint32_t domid, const char *rune, int debug, + int max_iters, int max_factor, int abort_if_busy, const char *override_config_file) { pid_t child = -1; @@ -3369,7 +3370,12 @@ static void migrate_domain(uint32_t domi char *away_domname; char rc_buf; uint8_t *config_data; - int config_len, flags = LIBXL_SUSPEND_LIVE; + int config_len; + libxl_domain_suspend_properties props = { + .flags = LIBXL_SUSPEND_LIVE, + .max_iters = max_iters, + .max_factor = max_factor, + }; save_domain_core_begin(domid, override_config_file, &config_data, &config_len); @@ -3388,8 +3394,11 @@ static void migrate_domain(uint32_t domi xtl_stdiostream_adjust_flags(logger, XTL_STDIOSTREAM_HIDE_PROGRESS, 0); if (debug) - flags |= LIBXL_SUSPEND_DEBUG; - rc = libxl_domain_suspend(ctx, domid, send_fd, flags, NULL); + props.flags |= LIBXL_SUSPEND_DEBUG; + if (abort_if_busy) + props.flags |= LIBXL_SUSPEND_ABORT_IF_BUSY; + + rc = libxl_domain_suspend(ctx, domid, send_fd, &props, NULL); if (rc) { fprintf(stderr, "migration sender: libxl_domain_suspend failed" " (rc=%d)\n", rc); @@ -3782,13 +3791,17 @@ int main_migrate(int argc, char **argv) char *rune = NULL; char *host; int opt, daemonize = 1, monitor = 1, debug = 0; + int max_iters = 0, max_factor = 0, abort_if_busy = 0; static struct option opts[] = { {"debug", 0, 0, 0x100}, + {"max_iters", 1, 0, 0x101}, + {"max_factor", 1, 0, 0x102}, + {"abort_if_busy", 0, 0, 0x103}, COMMON_LONG_OPTS, {0, 0, 0, 0} }; - SWITCH_FOREACH_OPT(opt, "FC:s:e", opts, "migrate", 2) { + SWITCH_FOREACH_OPT(opt, "FC:s:eM:m:A", opts, "migrate", 2) { case 'C': config_filename = optarg; break; @@ -3805,6 +3818,15 @@ int main_migrate(int argc, char **argv) case 0x100: debug = 1; break; + case 0x101: + max_iters = atoi(optarg); + break; + case 0x102: + max_factor = atoi(optarg); + break; + case 0x103: + abort_if_busy = 1; + break; } domid = find_domain(argv[optind]); @@ -3820,7 +3842,7 @@ int main_migrate(int argc, char **argv) return 1; } - migrate_domain(domid, rune, debug, config_filename); + migrate_domain(domid, rune, debug, max_iters, max_factor, abort_if_busy, config_filename); return 0; } diff -r b3ed5c4b5078 -r a897276ed24d tools/libxl/xl_cmdtable.c --- a/tools/libxl/xl_cmdtable.c +++ b/tools/libxl/xl_cmdtable.c @@ -147,14 +147,18 @@ struct cmd_spec cmd_table[] = { &main_migrate, 0, 1, "Migrate a domain to another host", "[options] <Domain> <host>", - "-h Print this help.\n" - "-C <config> Send <config> instead of config file from creation.\n" - "-s <sshcommand> Use <sshcommand> instead of ssh. String will be passed\n" - " to sh. If empty, run <host> instead of ssh <host> xl\n" - " migrate-receive [-d -e]\n" - "-e Do not wait in the background (on <host>) for the death\n" - " of the domain.\n" - "--debug Print huge (!) amount of debug during the migration process." + "-h Print this help.\n" + "-C <config> Send <config> instead of config file from creation.\n" + "-s <sshcommand> Use <sshcommand> instead of ssh. String will be passed\n" + " to sh. If empty, run <host> instead of ssh <host> xl\n" + " migrate-receive [-d -e]\n" + "-e Do not wait in the background (on <host>) for the death\n" + " of the domain.\n" + "--debug Print huge (!) amount of debug during the migration process.\n" + "--max_iters <number> Number of iterations before final suspend (default: 30)\n" + "--max_factor <factor> Max amount of memory to transfer before final suspend (default: 3*RAM).\n" + "--abort_if_busy Abort migration instead of doing final suspend, if number\n" + " of iterations or amount of transfered memory is exceeded." }, { "dump-core", &main_dump_core, 0, 1, diff -r b3ed5c4b5078 -r a897276ed24d tools/python/xen/xend/XendCheckpoint.py --- a/tools/python/xen/xend/XendCheckpoint.py +++ b/tools/python/xen/xend/XendCheckpoint.py @@ -118,9 +118,19 @@ def save(fd, dominfo, network, live, dst # enabled. Passing "0" simply uses the defaults compiled into # libxenguest; see the comments and/or code in xc_linux_save() for # more information. + max_iters = dominfo.info.get('max_iters', "0") + max_factor = dominfo.info.get('max_factor', "0") + abort_if_busy = dominfo.info.get('abort_if_busy', "0") + if max_iters == "None": + max_iters = "0" + if max_factor == "None": + max_factor = "0" + if abort_if_busy == "None": + abort_if_busy = "0" cmd = [xen.util.auxbin.pathTo(XC_SAVE), str(fd), - str(dominfo.getDomid()), "0", "0", - str(int(live) | (int(hvm) << 2)) ] + str(dominfo.getDomid()), + max_iters, max_factor, + str( int(live) | (int(hvm) << 2) | (int(abort_if_busy) << 5) ) ] log.debug("[xc_save]: %s", string.join(cmd)) def saveInputHandler(line, tochild): diff -r b3ed5c4b5078 -r a897276ed24d tools/python/xen/xend/XendDomain.py --- a/tools/python/xen/xend/XendDomain.py +++ b/tools/python/xen/xend/XendDomain.py @@ -1832,6 +1832,18 @@ class XendDomain: log.exception(ex) raise XendError(str(ex)) + def domain_migrate_constraints_set(self, domid, max_iters, max_factor, abort_if_busy): + """Set the Migrate Constraints of this domain. + @param domid: Domain ID or Name + @param max_iters: Number of iterations before final suspend + @param max_factor: Max amount of memory to transfer before final suspend + @param abort_if_busy: Abort migration instead of doing final suspend + """ + dominfo = self.domain_lookup_nr(domid) + if not dominfo: + raise XendInvalidDomain(str(domid)) + dominfo.setMigrateConstraints(max_iters, max_factor, abort_if_busy) + def domain_maxmem_set(self, domid, mem): """Set the memory limit for a domain. diff -r b3ed5c4b5078 -r a897276ed24d tools/python/xen/xend/XendDomainInfo.py --- a/tools/python/xen/xend/XendDomainInfo.py +++ b/tools/python/xen/xend/XendDomainInfo.py @@ -1459,6 +1459,18 @@ class XendDomainInfo: pci_conf = self.info['devices'][dev_uuid][1] return map(pci_dict_to_bdf_str, pci_conf['devs']) + def setMigrateConstraints(self, max_iters, max_factor, abort_if_busy): + """Set the Migrate Constraints of this domain. + @param max_iters: Number of iterations before final suspend + @param max_factor: Max amount of memory to transfer before final suspend + @param abort_if_busy: Abort migration instead of doing final suspend + """ + log.debug("Setting migration constraints of domain %s (%s) to '%s' '%s' '%s'.", + self.info['name_label'], str(self.domid), max_iters, max_factor, abort_if_busy) + self.info['max_iters'] = str(max_iters) + self.info['max_factor'] = str(max_factor) + self.info['abort_if_busy'] = str(abort_if_busy) + def setMemoryTarget(self, target): """Set the memory target of this domain. @param target: In MiB. diff -r b3ed5c4b5078 -r a897276ed24d tools/python/xen/xm/migrate.py --- a/tools/python/xen/xm/migrate.py +++ b/tools/python/xen/xm/migrate.py @@ -55,6 +55,18 @@ gopts.opt('change_home_server', short='c fn=set_true, default=0, use="Change home server for managed domains.") +gopts.opt('max_iters',val='max_iters', + fn=set_int, default=0, + use="Number of iterations before final suspend (default: 30).") + +gopts.opt('max_factor', val='max_factor', + fn=set_int, default=0, + use="Max amount of memory to transfer before final suspend (default: 3*RAM).") + +gopts.opt('abort_if_busy', + fn=set_true, default=0, + use="Abort migration instead of doing final suspend.") + def help(): return str(gopts) @@ -80,6 +92,10 @@ def main(argv): server.xenapi.VM.migrate(vm_ref, dst, bool(opts.vals.live), other_config) else: + server.xend.domain.migrate_constraints_set(dom, + opts.vals.max_iters, + opts.vals.max_factor, + opts.vals.abort_if_busy) server.xend.domain.migrate(dom, dst, opts.vals.live, opts.vals.port, opts.vals.node, _______________________________________________ Xen-devel mailing list Xen-devel@xxxxxxxxxxxxx http://lists.xen.org/xen-devel
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |