ChenSammi commented on code in PR #3982:
URL: https://github.com/apache/ozone/pull/3982#discussion_r1046629992
##########
hadoop-hdds/framework/src/main/java/org/apache/hadoop/hdds/security/x509/certificate/client/DefaultCertificateClient.java:
##########
@@ -1095,4 +1089,274 @@ public synchronized void close() throws IOException {
clientKeyStoresFactory.destroy();
}
}
+
+ /**
+ * Check how much time before certificate will enter expiry grace period.
+ * @return Duration, time before certificate enters the grace
+ * period defined by "hdds.x509.renew.grace.duration"
+ */
+ public Duration timeBeforeExpiryGracePeriod(String certId)
+ throws CertificateException {
+ X509Certificate cert = getCertificate(certId);
+ Duration gracePeriod = securityConfig.getRenewalGracePeriod();
+ Date expireDate = cert.getNotAfter();
+ LocalDateTime gracePeriodStart = expireDate.toInstant()
+ .atZone(ZoneId.systemDefault()).toLocalDateTime().minus(gracePeriod);
+ LocalDateTime currentTime = LocalDateTime.now();
+ if (gracePeriodStart.isBefore(currentTime)) {
+ // Cert is already in grace period time.
+ return Duration.ZERO;
+ } else {
+ return Duration.between(currentTime, gracePeriodStart);
+ }
+ }
+
+ public String renewAndStoreKeyAndCertificate(boolean force)
+ throws CertificateException {
+ if (isRenewing.compareAndSet(false, true)) {
+ try {
+ if (!force) {
+ synchronized (this) {
+ Preconditions.checkArgument(
+ timeBeforeExpiryGracePeriod(certSerialId).isZero());
+ }
+ }
+ String newKeyPath = securityConfig.getKeyLocation(component)
+ .toString() + HDDS_NEW_KEY_CERT_DIR_NAME_SUFFIX;
+ String newCertPath = securityConfig.getCertificateLocation(component)
+ .toString() + HDDS_NEW_KEY_CERT_DIR_NAME_SUFFIX;
+ File newKeyDir = new File(newKeyPath);
+ File newCertDir = new File(newCertPath);
+
+ try {
+ FileUtils.deleteDirectory(newKeyDir);
+ FileUtils.deleteDirectory(newCertDir);
+ } catch (IOException e) {
+ throw new CertificateException("Error while deleting " + newKeyPath +
+ " or " + newCertPath + " directories to cleanup certificate " +
+ " storage. ", e, RENEW_ERROR);
+ }
+
+ try {
+ Files.createDirectories(newKeyDir.toPath());
+ Files.createDirectories(newCertDir.toPath());
+ } catch (IOException e) {
+ throw new CertificateException("Error while creating " + newKeyPath +
+ " or " + newCertPath + " directories for certificate storage.",
+ e, RENEW_ERROR);
+ }
+
+ // cleanup backup directory
+ cleanBackupDir();
+
+ // Generate key
+ KeyCodec newKeyCodec = new KeyCodec(securityConfig,
newKeyDir.toPath());
+ KeyPair newKeyPair;
+ try {
+ newKeyPair = createKeyPair(newKeyCodec);
+ } catch (CertificateException e) {
+ throw new CertificateException("Error while creating new key pair.",
+ e, RENEW_ERROR);
+ }
+
+ // Get certificate signed
+ String dnCertSerialId;
+ try {
+ CertificateSignRequest.Builder csrBuilder =
getCSRBuilder(newKeyPair);
+ dnCertSerialId = signAndStoreCertificate(csrBuilder.build(),
+ Paths.get(newCertPath));
+ } catch (Exception e) {
+ throw new CertificateException("Error while signing and storing new"
+
+ " certificates.", e, RENEW_ERROR);
+ }
+
+ // switch Key and Certs directory on disk
+ File currentKeyDir = new File(
+ securityConfig.getKeyLocation(component).toString());
+ File currentCertDir = new File(
+ securityConfig.getCertificateLocation(component).toString());
+ File backupKeyDir = new File(
+ securityConfig.getKeyLocation(component).toString() +
+ HDDS_BACKUP_KEY_CERT_DIR_NAME_SUFFIX);
+ File backupCertDir = new File(
+ securityConfig.getCertificateLocation(component).toString() +
+ HDDS_BACKUP_KEY_CERT_DIR_NAME_SUFFIX);
+
+ if (!currentKeyDir.renameTo(backupKeyDir)) {
+ // Cannot rename current key dir to the backup dir
+ throw new CertificateException("Failed to rename " +
+ currentKeyDir.getAbsolutePath() +
+ " to " + backupKeyDir.getAbsolutePath() + " during " +
+ "certificate renew.", RENEW_ERROR);
+ }
+ if (!currentCertDir.renameTo(backupCertDir)) {
+ // Cannot rename current cert dir to the backup dir
+ rollbackDir(currentKeyDir, currentCertDir, newKeyDir, newCertDir,
+ backupKeyDir, backupCertDir, "step-1");
+ throw new CertificateException("Failed to rename " +
+ currentCertDir.getAbsolutePath() +
+ " to " + backupCertDir.getAbsolutePath() + " during " +
+ "certificate renew.", RENEW_ERROR);
+ }
+
+ if (!newKeyDir.renameTo(currentKeyDir)) {
+ // Cannot rename new dir as the current dir
+ String msg = "Failed to rename " + newKeyDir.getAbsolutePath() +
+ " to " + currentKeyDir.getAbsolutePath() +
+ " during certificate renew.";
+ // rollback
+ rollbackDir(currentKeyDir, currentCertDir, newKeyDir, newCertDir,
+ backupKeyDir, backupCertDir, "step-2");
+ throw new CertificateException(msg, RENEW_ERROR);
+ }
+
+ if (!newCertDir.renameTo(currentCertDir)) {
+ // Cannot rename new dir as the current dir
+ String msg = "Failed to rename " + newCertDir.getAbsolutePath() +
+ " to " + currentCertDir.getAbsolutePath() +
+ " during certificate renew.";
+ // rollback
+ rollbackDir(currentKeyDir, currentCertDir, newKeyDir, newCertDir,
+ backupKeyDir, backupCertDir, "step-2");
+ throw new CertificateException(msg, RENEW_ERROR);
+ }
+
+ // Delete backup dir on next DN startup
+ getLogger().info("Successful renew key and certificate." +
+ " New certificate {}.", dnCertSerialId);
+ return dnCertSerialId;
+ } finally {
+ isRenewing.set(false);
+ }
+ }
+
+ throw new CertificateException("A renewAndStoreKeyAndCert process" +
+ " is running already.", RENEW_ERROR);
+ }
+
+ private void rollbackDir(File currentKeyDir, File currentCertDir,
+ File newKeyDir, File newCertDir, File backupKeyDir, File backupCertDir,
+ String step) throws CertificateException {
Review Comment:
@fapifta , one thing can improve is we may not need to move current back to
new if move new folder to current location failed.
Since we are doing the key and cert renew far before they expire, for some
recoverable failures, such as certificate sign request sent to SCM who cannot
response in time due to any unexpected reason, we should tolerate this case and
have renew retry later.
So I defined two CertificateException error code, one for renew error, one
for rollback error. Renew error means something is wrong with the key and cert
renew but it doesn't have bad impact to current key and certs, so the service
can continue run, for example, the above case.
Rollback error means there is something happened which requires admin manual
intervention, such as fail to move backup to current location. In this case,
the monitor task will terminate the service process.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
