//!
//! Defines the thread that does the actual reboot of an instance.
//!
//! @param[in] arg a transparent pointer to the argument passed to this thread handler
//!
//! @return Always return NULL
//!
static void *rebooting_thread(void *arg)
{
char *xml = NULL;
char resourceName[1][MAX_SENSOR_NAME_LEN] = { "" };
char resourceAlias[1][MAX_SENSOR_NAME_LEN] = { "" };
// ncInstance *instance = ((ncInstance *) arg);
ncInstance *instance = NULL;
struct nc_state_t *nc = NULL;
virDomainPtr dom = NULL;
virConnectPtr conn = NULL;
rebooting_thread_params *params = ((rebooting_thread_params *) arg);
instance = &(params->instance);
nc = &(params->nc);
LOGDEBUG("[%s] spawning rebooting thread\n", instance->instanceId);
if ((conn = lock_hypervisor_conn()) == NULL) {
LOGERROR("[%s] cannot connect to hypervisor to restart instance, giving up\n", instance->instanceId);
EUCA_FREE(params);
return NULL;
}
dom = virDomainLookupByName(conn, instance->instanceId);
if (dom == NULL) {
LOGERROR("[%s] cannot locate instance to reboot, giving up\n", instance->instanceId);
unlock_hypervisor_conn();
EUCA_FREE(params);
return NULL;
}
// obtain the most up-to-date XML for domain from libvirt
xml = virDomainGetXMLDesc(dom, 0);
if (xml == NULL) {
LOGERROR("[%s] cannot obtain metadata for instance to reboot, giving up\n", instance->instanceId);
virDomainFree(dom); // release libvirt resource
unlock_hypervisor_conn();
EUCA_FREE(params);
return NULL;
}
virDomainFree(dom); // release libvirt resource
unlock_hypervisor_conn();
// try shutdown first, then kill it if uncooperative
if (shutdown_then_destroy_domain(instance->instanceId, TRUE) != EUCA_OK) {
LOGERROR("[%s] failed to shutdown and destroy the instance to reboot, giving up\n", instance->instanceId);
EUCA_FREE(params);
return NULL;
}
// Add a shift to values of three of the metrics: ones that
// drop back to zero after a reboot. The shift, which is based
// on the latest value, ensures that values sent upstream do
// not go backwards .
sensor_shift_metric(instance->instanceId, "CPUUtilization");
sensor_shift_metric(instance->instanceId, "NetworkIn");
sensor_shift_metric(instance->instanceId, "NetworkOut");
if ((conn = lock_hypervisor_conn()) == NULL) {
LOGERROR("[%s] cannot connect to hypervisor to restart instance, giving up\n", instance->instanceId);
EUCA_FREE(params);
return NULL;
}
// domain is now shut down, create a new one with the same XML
LOGINFO("[%s] rebooting\n", instance->instanceId);
if (!strcmp(nc->pEucaNet->sMode, NETMODE_VPCMIDO)) {
// need to sleep to allow midolman to update the VM interface
sleep(10);
}
dom = virDomainCreateLinux(conn, xml, 0);
if (dom == NULL) {
LOGERROR("[%s] failed to restart instance\n", instance->instanceId);
change_state(instance, SHUTOFF);
} else {
euca_strncpy(resourceName[0], instance->instanceId, MAX_SENSOR_NAME_LEN);
sensor_refresh_resources(resourceName, resourceAlias, 1); // refresh stats so we set base value accurately
virDomainFree(dom);
if (!strcmp(nc->pEucaNet->sMode, NETMODE_VPCMIDO)) {
char iface[16], cmd[EUCA_MAX_PATH], obuf[256], ebuf[256], sPath[EUCA_MAX_PATH];
int rc;
snprintf(iface, 16, "vn_%s", instance->instanceId);
// If this device does not have a 'brport' path, this isn't a bridge device
snprintf(sPath, EUCA_MAX_PATH, "/sys/class/net/%s/brport/", iface);
if (!check_directory(sPath)) {
LOGDEBUG("[%s] removing instance interface %s from host bridge\n", instance->instanceId, iface);
snprintf(cmd, EUCA_MAX_PATH, "%s brctl delif %s %s", nc->rootwrap_cmd_path, instance->params.guestNicDeviceName, iface);
rc = timeshell(cmd, obuf, ebuf, 256, 10);
if (rc) {
LOGERROR("unable to remove instance interface from bridge after launch: instance will not be able to connect to midonet (will not connect to network): check bridge/libvirt/kvm health\n");
}
}
// Repeat process for secondary interfaces as well
for (int i=0; i < EUCA_MAX_NICS; i++) {
if (strlen(instance->secNetCfgs[i].interfaceId) == 0)
continue;
snprintf(iface, 16, "vn_%s", instance->secNetCfgs[i].interfaceId);
// If this device does not have a 'brport' path, this isn't a bridge device
snprintf(sPath, EUCA_MAX_PATH, "/sys/class/net/%s/brport/", iface);
if (!check_directory(sPath)) {
LOGDEBUG("[%s] removing instance interface %s from host bridge\n", instance->instanceId, iface);
snprintf(cmd, EUCA_MAX_PATH, "%s brctl delif %s %s", nc->rootwrap_cmd_path, instance->params.guestNicDeviceName, iface);
rc = timeshell(cmd, obuf, ebuf, 256, 10);
if (rc) {
LOGERROR("unable to remove instance interface from bridge after launch: instance will not be able to connect to midonet (will not connect to network): check bridge/libvirt/kvm health\n");
}
}
}
}
}
EUCA_FREE(xml);
unlock_hypervisor_conn();
unset_corrid(get_corrid());
EUCA_FREE(params);
return NULL;
}
//!
//! Defines the thread that does the actual reboot of an instance.
//!
//! @param[in] arg a transparent pointer to the argument passed to this thread handler
//!
//! @return Always return NULL
//!
static void *rebooting_thread(void *arg)
{
#define REATTACH_RETRIES 3
int i = 0;
int err = 0;
int error = 0;
int rc = 0;
int log_level_for_devstring = EUCATRACE;
char *xml = NULL;
char *remoteDevStr = NULL;
char path[MAX_PATH] = "";
char lpath[MAX_PATH] = "";
char resourceName[1][MAX_SENSOR_NAME_LEN] = { {0} };
char resourceAlias[1][MAX_SENSOR_NAME_LEN] = { {0} };
ncVolume *volume = NULL;
ncInstance *instance = ((ncInstance *) arg);
virDomainPtr dom = NULL;
virConnectPtr *conn = NULL;
logprintfl(EUCADEBUG, "[%s] spawning rebooting thread\n", instance->instanceId);
if ((xml = file2str(instance->libvirtFilePath)) == NULL) {
logprintfl(EUCAERROR, "[%s] cannot obtain instance XML file %s\n", instance->instanceId, instance->libvirtFilePath);
return NULL;
}
if ((conn = check_hypervisor_conn()) == NULL) {
logprintfl(EUCAERROR, "[%s] cannot restart instance %s, abandoning it\n", instance->instanceId, instance->instanceId);
change_state(instance, SHUTOFF);
EUCA_FREE(xml);
return NULL;
}
sem_p(hyp_sem);
{
dom = virDomainLookupByName(*conn, instance->instanceId);
}
sem_v(hyp_sem);
if (dom == NULL) {
EUCA_FREE(xml);
return NULL;
}
sem_p(hyp_sem);
{
// for KVM, must stop and restart the instance
logprintfl(EUCADEBUG, "[%s] destroying domain\n", instance->instanceId);
error = virDomainDestroy(dom); // @todo change to Shutdown? is this synchronous?
virDomainFree(dom);
}
sem_v(hyp_sem);
if (error) {
EUCA_FREE(xml);
return NULL;
}
// Add a shift to values of three of the metrics: ones that
// drop back to zero after a reboot. The shift, which is based
// on the latest value, ensures that values sent upstream do
// not go backwards .
sensor_shift_metric(instance->instanceId, "CPUUtilization");
sensor_shift_metric(instance->instanceId, "NetworkIn");
sensor_shift_metric(instance->instanceId, "NetworkOut");
// domain is now shut down, create a new one with the same XML
sem_p(hyp_sem);
{
logprintfl(EUCAINFO, "[%s] rebooting\n", instance->instanceId);
dom = virDomainCreateLinux(*conn, xml, 0);
}
sem_v(hyp_sem);
EUCA_FREE(xml);
euca_strncpy(resourceName[0], instance->instanceId, MAX_SENSOR_NAME_LEN);
sensor_refresh_resources(resourceName, resourceAlias, 1); // refresh stats so we set base value accurately
// re-attach each volume previously attached
for (i = 0; i < EUCA_MAX_VOLUMES; ++i) {
volume = &instance->volumes[i];
if (strcmp(volume->stateName, VOL_STATE_ATTACHED) && strcmp(volume->stateName, VOL_STATE_ATTACHING))
continue; // skip the entry unless attached or attaching
logprintfl(EUCADEBUG, "[%s] volumes [%d] = '%s'\n", instance->instanceId, i, volume->stateName);
// get credentials, decrypt them
remoteDevStr = get_iscsi_target(volume->remoteDev);
if (!remoteDevStr || !strstr(remoteDevStr, "/dev")) {
logprintfl(EUCAERROR, "[%s] failed to get local name of host iscsi device when re-attaching\n", instance->instanceId);
rc = 1;
} else {
// set the path
snprintf(path, sizeof(path), EUCALYPTUS_VOLUME_XML_PATH_FORMAT, instance->instancePath, volume->volumeId); // vol-XXX.xml
snprintf(lpath, sizeof(lpath), EUCALYPTUS_VOLUME_LIBVIRT_XML_PATH_FORMAT, instance->instancePath, volume->volumeId); // vol-XXX-libvirt.xml
// read in libvirt XML, which may have been modified by the hook above
if ((xml = file2str(lpath)) == NULL) {
logprintfl(EUCAERROR, "[%s][%s] failed to read volume XML from %s\n", instance->instanceId, volume->volumeId, lpath);
rc = 1;
}
}
EUCA_FREE(remoteDevStr);
if (!rc) {
// zhill - wrap with retry in case libvirt is dumb.
err = 0;
for (i = 1; i < REATTACH_RETRIES; i++) {
// protect libvirt calls because we've seen problems during concurrent libvirt invocations
sem_p(hyp_sem);
{
err = virDomainAttachDevice(dom, xml);
}
sem_v(hyp_sem);
if (err) {
logprintfl(EUCAERROR, "[%s][%s] failed to reattach volume (attempt %d of %d)\n", instance->instanceId, volume->volumeId, i,
REATTACH_RETRIES);
logprintfl(EUCADEBUG, "[%s][%s] error from virDomainAttachDevice: %d xml: %s\n", instance->instanceId, volume->volumeId, err,
xml);
sleep(3); // sleep a bit and retry
} else {
logprintfl(EUCAINFO, "[%s][%s] volume reattached as '%s'\n", instance->instanceId, volume->volumeId, volume->localDevReal);
break;
}
}
log_level_for_devstring = EUCATRACE;
if (err)
log_level_for_devstring = EUCADEBUG;
logprintfl(log_level_for_devstring, "[%s][%s] remote device string: %s\n", instance->instanceId, volume->volumeId, volume->remoteDev);
}
EUCA_FREE(xml);
}
if (dom == NULL) {
logprintfl(EUCAERROR, "[%s] failed to restart instance\n", instance->instanceId);
change_state(instance, SHUTOFF);
return NULL;
}
sem_p(hyp_sem);
{
virDomainFree(dom);
}
sem_v(hyp_sem);
return NULL;
#undef REATTACH_RETRIES
}
//!
//! Defines the thread that does the actual reboot of an instance.
//!
//! @param[in] arg a transparent pointer to the argument passed to this thread handler
//!
//! @return Always return NULL
//!
static void *rebooting_thread(void *arg)
{
char *xml = NULL;
char resourceName[1][MAX_SENSOR_NAME_LEN] = { "" };
char resourceAlias[1][MAX_SENSOR_NAME_LEN] = { "" };
ncInstance *instance = ((ncInstance *) arg);
virDomainPtr dom = NULL;
virConnectPtr conn = NULL;
LOGDEBUG("[%s] spawning rebooting thread\n", instance->instanceId);
if ((conn = lock_hypervisor_conn()) == NULL) {
LOGERROR("[%s] cannot connect to hypervisor to restart instance, giving up\n", instance->instanceId);
return NULL;
}
dom = virDomainLookupByName(conn, instance->instanceId);
if (dom == NULL) {
LOGERROR("[%s] cannot locate instance to reboot, giving up\n", instance->instanceId);
unlock_hypervisor_conn();
return NULL;
}
// obtain the most up-to-date XML for domain from libvirt
xml = virDomainGetXMLDesc(dom, 0);
if (xml == NULL) {
LOGERROR("[%s] cannot obtain metadata for instance to reboot, giving up\n", instance->instanceId);
virDomainFree(dom); // release libvirt resource
unlock_hypervisor_conn();
return NULL;
}
virDomainFree(dom); // release libvirt resource
unlock_hypervisor_conn();
// try shutdown first, then kill it if uncooperative
if (shutdown_then_destroy_domain(instance->instanceId, TRUE) != EUCA_OK) {
LOGERROR("[%s] failed to shutdown and destroy the instance to reboot, giving up\n", instance->instanceId);
return NULL;
}
// Add a shift to values of three of the metrics: ones that
// drop back to zero after a reboot. The shift, which is based
// on the latest value, ensures that values sent upstream do
// not go backwards .
sensor_shift_metric(instance->instanceId, "CPUUtilization");
sensor_shift_metric(instance->instanceId, "NetworkIn");
sensor_shift_metric(instance->instanceId, "NetworkOut");
if ((conn = lock_hypervisor_conn()) == NULL) {
LOGERROR("[%s] cannot connect to hypervisor to restart instance, giving up\n", instance->instanceId);
return NULL;
}
// domain is now shut down, create a new one with the same XML
LOGINFO("[%s] rebooting\n", instance->instanceId);
dom = virDomainCreateLinux(conn, xml, 0);
if (dom == NULL) {
LOGERROR("[%s] failed to restart instance\n", instance->instanceId);
change_state(instance, SHUTOFF);
} else {
euca_strncpy(resourceName[0], instance->instanceId, MAX_SENSOR_NAME_LEN);
sensor_refresh_resources(resourceName, resourceAlias, 1); // refresh stats so we set base value accurately
virDomainFree(dom);
}
EUCA_FREE(xml);
unlock_hypervisor_conn();
unset_corrid(get_corrid());
return NULL;
}
