Copilot commented on code in PR #14160: URL: https://github.com/apache/cloudstack/pull/14160#discussion_r4011645205
########## plugins/storage/object/seaweedfs/src/main/java/org/apache/cloudstack/storage/datastore/driver/SeaweedFSObjectStoreDriverImpl.java: ########## @@ -0,0 +1,838 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ +// SPDX-License-Identifier: Apache-2.0 +package org.apache.cloudstack.storage.datastore.driver; + +import java.util.ArrayList; +import java.util.HashMap; +import java.util.List; +import java.util.Map; + +import javax.inject.Inject; + +import org.apache.cloudstack.engine.subsystem.api.storage.DataStore; +import org.apache.cloudstack.storage.datastore.db.ObjectStoreDao; +import org.apache.cloudstack.storage.datastore.db.ObjectStoreDetailsDao; +import org.apache.cloudstack.storage.datastore.db.ObjectStoreVO; +import org.apache.cloudstack.storage.datastore.util.SeaweedFSObjectStoreUtil; +import org.apache.cloudstack.storage.object.BaseObjectStoreDriverImpl; +import org.apache.cloudstack.storage.object.Bucket; +import org.apache.cloudstack.storage.object.BucketObject; + +import com.amazonaws.AmazonClientException; +import com.amazonaws.services.identitymanagement.AmazonIdentityManagement; +import com.amazonaws.services.identitymanagement.model.AccessKey; +import com.amazonaws.services.identitymanagement.model.AccessKeyMetadata; +import com.amazonaws.services.identitymanagement.model.CreateAccessKeyRequest; +import com.amazonaws.services.identitymanagement.model.CreateAccessKeyResult; +import com.amazonaws.services.identitymanagement.model.CreateUserRequest; +import com.amazonaws.services.identitymanagement.model.DeleteAccessKeyRequest; +import com.amazonaws.services.identitymanagement.model.EntityAlreadyExistsException; +import com.amazonaws.services.identitymanagement.model.ListAccessKeysRequest; +import com.amazonaws.services.identitymanagement.model.PutUserPolicyRequest; +import com.amazonaws.services.s3.AmazonS3; +import com.amazonaws.services.s3.model.AccessControlList; +import com.amazonaws.services.s3.model.BucketPolicy; +import com.amazonaws.services.s3.model.BucketVersioningConfiguration; +import com.amazonaws.services.s3.model.CreateBucketRequest; +import com.amazonaws.services.s3.model.DeleteBucketPolicyRequest; +import com.amazonaws.services.s3.model.BucketCrossOriginConfiguration; +import com.amazonaws.services.s3.model.CORSRule; +import com.amazonaws.services.s3.model.GetBucketPolicyRequest; +import com.amazonaws.services.s3.model.SSEAlgorithm; +import com.amazonaws.services.s3.model.ServerSideEncryptionByDefault; +import com.amazonaws.services.s3.model.ServerSideEncryptionConfiguration; +import com.amazonaws.services.s3.model.ServerSideEncryptionRule; +import com.amazonaws.services.s3.model.SetBucketCrossOriginConfigurationRequest; +import com.amazonaws.services.s3.model.SetBucketEncryptionRequest; +import com.amazonaws.services.s3.model.SetBucketVersioningConfigurationRequest; +import com.cloud.agent.api.to.BucketTO; +import com.cloud.agent.api.to.DataStoreTO; +import com.cloud.storage.BucketVO; +import com.cloud.storage.dao.BucketDao; +import com.cloud.user.Account; +import com.cloud.user.AccountDetailsDao; +import com.cloud.user.dao.AccountDao; +import com.cloud.utils.db.GlobalLock; +import com.cloud.utils.exception.CloudRuntimeException; + +/** + * SeaweedFS object store driver. + * + * Bucket operations use the AWS S3 SDK v1 (path-style access, endpoint-pinned). + * User/credential management uses the AWS IAM SDK v1, since SeaweedFS exposes a + * standard AWS IAM-compatible API. No proprietary admin client is needed. + * + * Modeled on CloudianHyperStoreObjectStoreDriverImpl, which uses the same + * S3 + IAM SDK pair. + */ +public class SeaweedFSObjectStoreDriverImpl extends BaseObjectStoreDriverImpl { + + @Inject + AccountDao _accountDao; + + @Inject + AccountDetailsDao _accountDetailsDao; + + @Inject + ObjectStoreDao _storeDao; + + @Inject + BucketDao _bucketDao; + + @Inject + ObjectStoreDetailsDao _storeDetailsDao; + + private static final String ACS_PREFIX = "acs"; + + /** + * DB-backed global lock name prefix for serializing IAM provisioning and + * policy refreshes per store+account. Uses {@link GlobalLock} so the + * critical section is serialized across management servers in a + * clustered deployment, not just within a single JVM. + */ + private static final String IAM_LOCK_PREFIX = "seaweedfs.iam."; + + private static String getIamLockName(long storeId, long accountId) { + return IAM_LOCK_PREFIX + storeId + "." + accountId; + } + + /** + * Acquire a DB-backed global lock for IAM operations on the given + * store+account. Returns a {@link GlobalLock} that the caller must + * {@link GlobalLock#unlock()} in a {@code finally} block, or {@code null} + * if the lock could not be acquired within the timeout. + * + * <p>Protected so tests can override with a no-op lock (the DB-backed + * {@link GlobalLock} requires a real transaction context). + */ + protected GlobalLock acquireIamLock(long storeId, long accountId) { + GlobalLock lock = GlobalLock.getInternLock(getIamLockName(storeId, accountId)); + if (!lock.lock(300)) { + logger.warn("Failed to acquire IAM lock for store {} account {}", storeId, accountId); + lock.releaseRef(); + return null; + } + return lock; + } + + @Override + public DataStoreTO getStoreTO(DataStore store) { + return null; + } + + /** + * Get the SeaweedFS IAM user name for the given CloudStack account and + * store. The store ID is included so that two CloudStack pools pointing + * at the same SeaweedFS IAM service do not collide on the same + * {@code acs-<uuid>} user and overwrite each other's policy and access + * keys. + */ + protected String getUserNameForAccount(Account account, long storeId) { + return String.format("%s-%d-%s", ACS_PREFIX, storeId, account.getUuid()); + } + + /** + * Create the IAM user for the CloudStack account if it doesn't exist, + * attach the restricted S3 policy, and ensure the account has a usable + * IAM access key persisted in its account details. + * + * <p>If a previously stored access key is still present in IAM, it is + * reused rather than rotated. A new key is only created when no stored + * key exists or the stored key is no longer found in IAM; in the latter + * case any unmanaged (leftover) keys for the user are deleted first to + * avoid hitting IAM access-key limits. This keeps bucket records that + * reference the stored credentials valid across repeated calls. + * + * @return true if the user exists or was created, false on failure. + */ + @Override + public boolean createUser(long accountId, long storeId) { + Account account = _accountDao.findById(accountId); + if (account == null) { + logger.error("Account {} not found", accountId); + return false; + } + String userName = getUserNameForAccount(account, storeId); + AmazonIdentityManagement iamClient = getIAMClient(storeId); + + // Serialize per store+account across management servers so two + // concurrent bucket requests do not both rotate credentials and leave + // bucket rows with mismatched key pairs. + GlobalLock lock = acquireIamLock(storeId, accountId); + if (lock == null) { + return false; + } + try { + + // Create the IAM user if it doesn't already exist + try { + iamClient.createUser(new CreateUserRequest(userName)); + logger.info("Created IAM user {} for account {}", userName, account.getAccountName()); + } catch (EntityAlreadyExistsException e) { + logger.debug("IAM user {} already exists", userName); + } + + // Attach a scoped IAM policy that allows access only to this + // account's own buckets (the tenant boundary). Refreshed whenever + // buckets are created or deleted. + updateAccountIAMPolicy(iamClient, storeId, accountId, null); + + // Reuse the stored access key only if both the access key id and the + // secret key are present and the key is still Active in IAM; otherwise + // create a replacement. + Map<String, String> details = _accountDetailsDao.findDetails(accountId); + String accessKeyDetailKey = SeaweedFSObjectStoreUtil.keyAccessKey(storeId); + String secretKeyDetailKey = SeaweedFSObjectStoreUtil.keySecretKey(storeId); + String storedAccessKeyId = details.get(accessKeyDetailKey); + String storedSecretKey = details.get(secretKeyDetailKey); + if (storedAccessKeyId != null && storedSecretKey != null + && iamAccessKeyExists(iamClient, userName, storedAccessKeyId)) { + logger.debug("Reusing existing IAM access key {} for user {}", storedAccessKeyId, userName); + return true; + } + + // The stored key is missing, inactive, or no longer in IAM. Clean up + // ALL keys (including the inactive stored one) before creating a + // replacement so we do not accumulate keys and hit IAM limits. + deleteUnmanagedAccessKeys(iamClient, userName, null); + + CreateAccessKeyResult result = iamClient.createAccessKey( + new CreateAccessKeyRequest().withUserName(userName)); + AccessKey key = result.getAccessKey(); + + // Update existing bucket records for this account/store with the new + // credentials BEFORE persisting the new key in account details. If a + // bucket update fails, the stored key remains the old one and a retry + // will re-enter the replacement path; if we persisted first, a retry + // would see the new stored key and return without repairing the + // remaining buckets. + updateAccountBucketCredentials(storeId, accountId, key); + + // Persist the credentials in the account details (namespaced by storeId) + details.put(accessKeyDetailKey, key.getAccessKeyId()); + details.put(secretKeyDetailKey, key.getSecretAccessKey()); + _accountDetailsDao.persist(accountId, details); + + logger.info("Created IAM credentials {} for user {}", key.getAccessKeyId(), userName); + return true; + } finally { + lock.unlock(); + lock.releaseRef(); + } + } + + /** + * Update the IAM credentials on all BucketVO rows for this store/account + * so previously created buckets reflect the new (rotated) key pair. + * Mirrors CloudianHyperStoreObjectStoreDriverImpl.updateAccountBucketCredentials. + */ + private void updateAccountBucketCredentials(long storeId, long accountId, AccessKey iamCredential) { + List<BucketVO> bucketList = _bucketDao.listByObjectStoreIdAndAccountId(storeId, accountId); + for (BucketVO bucketVO : bucketList) { + logger.info("Updating accountId={} bucket {} with new IAM credentials", accountId, bucketVO.getName()); + bucketVO.setAccessKey(iamCredential.getAccessKeyId()); + bucketVO.setSecretKey(iamCredential.getSecretAccessKey()); + _bucketDao.update(bucketVO.getId(), bucketVO); + } + } + + /** + * Refresh the per-account IAM user policy so it grants S3 access only to + * the account's current buckets (optionally excluding one, e.g. a bucket + * being deleted). This is the tenant boundary: each account's IAM + * credentials can only operate on that account's own buckets. + * + * @param iamClient the IAM client + * @param storeId the object store + * @param accountId the CloudStack account + * @param excludeBucket a bucket name to omit (e.g. a bucket being deleted), + * or null to include all of the account's buckets + */ + protected void updateAccountIAMPolicy(AmazonIdentityManagement iamClient, long storeId, long accountId, String excludeBucket) { + GlobalLock lock = acquireIamLock(storeId, accountId); + if (lock == null) { + throw new CloudRuntimeException("Failed to acquire IAM lock for store " + storeId + " account " + accountId); + } + try { + Account account = _accountDao.findById(accountId); + if (account == null) { + return; + } + String userName = getUserNameForAccount(account, storeId); + List<BucketVO> buckets = _bucketDao.listByObjectStoreIdAndAccountId(storeId, accountId); + List<String> bucketNames = new ArrayList<>(); + for (BucketVO bvo : buckets) { + if (excludeBucket != null && excludeBucket.equals(bvo.getName())) { + continue; + } + bucketNames.add(bvo.getName()); + } + String policy = SeaweedFSObjectStoreUtil.buildAccountIAMPolicy(bucketNames); + iamClient.putUserPolicy(new PutUserPolicyRequest(userName, + SeaweedFSObjectStoreUtil.IAM_USER_POLICY_NAME, policy)); + } finally { + lock.unlock(); + lock.releaseRef(); + } + } + + /** + * Check whether the given access key id is still listed and Active in IAM + * for the user. Listing failures are propagated rather than swallowed so + * a transient IAM outage does not send createUser into the replacement + * path (which would overwrite stored credentials and invalidate bucket + * records). + */ + private boolean iamAccessKeyExists(AmazonIdentityManagement iamClient, String userName, String accessKeyId) { + for (AccessKeyMetadata metadata : + iamClient.listAccessKeys(new ListAccessKeysRequest() + .withUserName(userName)).getAccessKeyMetadata()) { + if (accessKeyId.equals(metadata.getAccessKeyId())) { + return "Active".equalsIgnoreCase(metadata.getStatus()); + } + } + return false; + } + + /** + * Delete access keys for the user other than the (optionally) preserved + * key id. Used to clean up unmanaged leftover keys before creating a + * replacement so repeated calls do not hit IAM access-key limits. + */ + private void deleteUnmanagedAccessKeys(AmazonIdentityManagement iamClient, String userName, String preserveAccessKeyId) { + try { + for (AccessKeyMetadata metadata : + iamClient.listAccessKeys(new ListAccessKeysRequest() + .withUserName(userName)).getAccessKeyMetadata()) { + String keyId = metadata.getAccessKeyId(); + if (preserveAccessKeyId != null && preserveAccessKeyId.equals(keyId)) { + continue; + } + DeleteAccessKeyRequest deleteReq = + new DeleteAccessKeyRequest() + .withUserName(userName) + .withAccessKeyId(keyId); + logger.info("Deleting un-managed IAM access key {} for user {}", keyId, userName); + iamClient.deleteAccessKey(deleteReq); + } + } catch (AmazonClientException e) { + // Propagate so the caller does not proceed to create a replacement + // key while stale unmanaged keys remain (which could hit IAM key + // limits or leave orphaned credentials). + throw new CloudRuntimeException("Failed to clean up IAM access keys for user " + userName, e); + } + } + + @Override + public Bucket createBucket(Bucket bucket, boolean objectLock) { + String bucketName = bucket.getName(); + long storeId = bucket.getObjectStoreId(); + long accountId = bucket.getAccountId(); + + // Use the store's admin credentials to create the bucket + AmazonS3 s3client = getS3ClientByStoreId(storeId); + + // Check if the bucket already exists + try { + if (s3client.doesBucketExistV2(bucketName)) { + throw new CloudRuntimeException("Bucket already exists with name " + bucketName); + } + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + + // Create the bucket + try { + CreateBucketRequest request = new CreateBucketRequest(bucketName); + if (objectLock) { + request.setObjectLockEnabledForBucket(true); + } + s3client.createBucket(request); + } catch (AmazonClientException e) { + logger.error("Create bucket failed", e); + throw new CloudRuntimeException(e); + } + + // Step 2: update the bucket record with the account's IAM credentials, + // configure CORS, and refresh the IAM policy. If any of these fail, + // clean up the remote bucket so a retry does not find it already + // existing — mirroring the Cloudian createBucket pattern. + // + // Hold the IAM lock for the account so a concurrent createUser key + // rotation does not change the account credentials between reading + // them and writing the BucketVO, which would leave the bucket with + // a stale key pair. + GlobalLock iamLock = acquireIamLock(storeId, accountId); + if (iamLock == null) { + // The S3 bucket has already been created. Clean it up so a + // retry does not find it already existing, then throw. + try { + s3client.deleteBucket(bucketName); + } catch (AmazonClientException cleanupEx) { + logger.error("Failed to clean up bucket {} after IAM lock timeout", bucketName, cleanupEx); + } + throw new CloudRuntimeException("Failed to acquire IAM lock for store " + storeId + " account " + accountId); + } + try { + // Configure permissive CORS so the CloudStack S3 bucket browser + // (which performs list/upload/delete from the browser) can function. + // SeaweedFS supports the standard PutBucketCors operation. + configureBucketCORS(s3client, bucketName); + + Map<String, String> accountDetails = _accountDetailsDao.findDetails(accountId); + String accessKey = accountDetails.get(SeaweedFSObjectStoreUtil.keyAccessKey(storeId)); + String secretKey = accountDetails.get(SeaweedFSObjectStoreUtil.keySecretKey(storeId)); + if (accessKey == null || secretKey == null) { + throw new CloudRuntimeException("No IAM credentials found for account " + accountId + + " on store " + storeId + ". Run createUser before creating a bucket."); + } + + String s3Url = getS3Url(storeId); + BucketVO bucketVO = _bucketDao.findById(bucket.getId()); + bucketVO.setAccessKey(accessKey); + bucketVO.setSecretKey(secretKey); + bucketVO.setBucketURL(s3Url + "/" + bucketName); + _bucketDao.update(bucket.getId(), bucketVO); + + // Refresh the account's IAM policy to include the new bucket + AmazonIdentityManagement iamClient = getIAMClient(storeId); + updateAccountIAMPolicy(iamClient, storeId, accountId, null); + + // Return the updated BucketVO (not the stale input bucket) so + // BucketApiServiceImpl.createBucket does not overwrite the + // persisted credentials with the stale values. + return bucketVO; + } catch (Exception e) { + logger.error("Post-create bucket record update failed for {}; cleaning up remote bucket", bucketName, e); + try { + s3client.deleteBucket(bucketName); + logger.info("Cleanup of bucket {} succeeded", bucketName); + } catch (AmazonClientException cleanupEx) { + logger.error("Cleanup of bucket {} also failed", bucketName, cleanupEx); + } + // Revoke the IAM policy grant for the new bucket so the account's + // credentials cannot access a bucket that no longer exists. If the + // policy PUT succeeded before the DB update failed, the grant + // would otherwise persist and could be reused if another account + // later creates the same bucket name. + try { + AmazonIdentityManagement iamClient = getIAMClient(storeId); + updateAccountIAMPolicy(iamClient, storeId, accountId, bucketName); + } catch (Exception policyEx) { + logger.warn("Failed to revoke IAM policy for bucket {} after cleanup: {}", bucketName, policyEx.getMessage()); + } + throw new CloudRuntimeException(e); + } finally { + iamLock.unlock(); + iamLock.releaseRef(); + } + } + + /** + * Configure a permissive CORS policy on the bucket so the CloudStack + * S3 bucket browser (which performs list/upload/delete from the + * browser) can function. Mirrors the Cloudian configureBucketCORS. + */ + private void configureBucketCORS(AmazonS3 s3client, String bucketName) { + logger.debug("Configuring CORS for bucket {}", bucketName); + List<CORSRule> corsRules = new ArrayList<>(); + CORSRule allowAnyRule = new CORSRule().withId("AllowAny"); + allowAnyRule.setAllowedOrigins("*"); + allowAnyRule.setAllowedHeaders("*"); + allowAnyRule.setAllowedMethods( + CORSRule.AllowedMethods.HEAD, + CORSRule.AllowedMethods.GET, + CORSRule.AllowedMethods.PUT, + CORSRule.AllowedMethods.POST, + CORSRule.AllowedMethods.DELETE); + corsRules.add(allowAnyRule); + BucketCrossOriginConfiguration corsConfig = new BucketCrossOriginConfiguration(); + corsConfig.setRules(corsRules); + SetBucketCrossOriginConfigurationRequest corsRequest = new SetBucketCrossOriginConfigurationRequest(bucketName, corsConfig); + s3client.setBucketCrossOriginConfiguration(corsRequest); + logger.info("Successfully configured CORS for bucket {}", bucketName); + } + + @Override + public List<Bucket> listBuckets(long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + List<Bucket> bucketsList = new ArrayList<>(); + try { + List<com.amazonaws.services.s3.model.Bucket> s3Buckets = s3client.listBuckets(); + for (com.amazonaws.services.s3.model.Bucket s3Bucket : s3Buckets) { + Bucket bucket = new BucketObject(); + bucket.setName(s3Bucket.getName()); + bucketsList.add(bucket); + } + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + return bucketsList; + } + + @Override + public boolean deleteBucket(BucketTO bucket, long storeId) { + String bucketName = bucket.getName(); + long accountId = bucket.getAccountId(); + AmazonS3 s3client = getS3ClientByStoreId(storeId); + + // Delete the S3 bucket first. If this fails (non-empty bucket, + // transient error), the IAM policy is still intact so the user + // can empty the bucket and retry. + // + // If the bucket is already gone (e.g. from a previous partial + // failure where the S3 delete succeeded but the IAM policy refresh + // failed), skip the S3 delete so the retry is idempotent. + try { + if (s3client.doesBucketExistV2(bucketName)) { + s3client.deleteBucket(bucketName); + } + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + + // Remove the BucketVO row before refreshing the IAM policy so a + // concurrent createUser/createBucket policy rebuild (which reads the + // bucket list from the DB) cannot re-add the deleted bucket ARN + // between this exclusion and BucketApiServiceImpl's row removal. + // BucketApiServiceImpl's subsequent _bucketDao.remove is idempotent. + for (BucketVO bvo : _bucketDao.listByObjectStoreIdAndAccountId(storeId, accountId)) { + if (bucketName.equals(bvo.getName())) { + _bucketDao.remove(bvo.getId()); + break; + } + } + + // Refresh the account's IAM policy to drop the deleted bucket. + // Bucket names are reusable, so a stale grant would let the old + // account access a new tenant's bucket with the same name. This + // must succeed; if it fails, the caller sees the exception and can + // retry (the policy refresh is idempotent because the bucket is + // already gone from S3 and the DB, so excludeBucket is a no-op). + AmazonIdentityManagement iamClient = getIAMClient(storeId); + updateAccountIAMPolicy(iamClient, storeId, accountId, bucketName); + return true; + } + + @Override + public AccessControlList getBucketAcl(BucketTO bucket, long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + return s3client.getBucketAcl(bucket.getName()); + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + @Override + public void setBucketAcl(BucketTO bucket, AccessControlList acl, long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + s3client.setBucketAcl(bucket.getName(), acl); + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + @Override + public void setBucketPolicy(BucketTO bucket, String policy, long storeId) { + if ("private".equalsIgnoreCase(policy)) { + deleteBucketPolicy(bucket, storeId); + return; + } + + StringBuilder sb = new StringBuilder(); + sb.append("{\n"); + sb.append(" \"Version\": \"2012-10-17\",\n"); + sb.append(" \"Statement\": [\n"); + sb.append(" {\n"); + sb.append(" \"Sid\": \"PublicReadForObjects\",\n"); + sb.append(" \"Effect\": \"Allow\",\n"); + sb.append(" \"Principal\": \"*\",\n"); + sb.append(" \"Action\": \"s3:GetObject\",\n"); + sb.append(" \"Resource\": \"arn:aws:s3:::%s/*\"\n"); + sb.append(" }\n"); + sb.append(" ]\n"); + sb.append("}\n"); + + String jsonPolicy = String.format(sb.toString(), bucket.getName()); + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + s3client.setBucketPolicy(bucket.getName(), jsonPolicy); + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + @Override + public BucketPolicy getBucketPolicy(BucketTO bucket, long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + return s3client.getBucketPolicy(new GetBucketPolicyRequest(bucket.getName())); + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + @Override + public void deleteBucketPolicy(BucketTO bucket, long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + s3client.deleteBucketPolicy(new DeleteBucketPolicyRequest(bucket.getName())); + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + @Override + public boolean setBucketEncryption(BucketTO bucket, long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + SetBucketEncryptionRequest eRequest = new SetBucketEncryptionRequest(); + eRequest.setBucketName(bucket.getName()); + + ServerSideEncryptionByDefault sseByDefault = new ServerSideEncryptionByDefault(); + sseByDefault.setSSEAlgorithm(SSEAlgorithm.AES256.toString()); + + ServerSideEncryptionRule sseRule = new ServerSideEncryptionRule(); + sseRule.setApplyServerSideEncryptionByDefault(sseByDefault); + + List<ServerSideEncryptionRule> sseRules = new ArrayList<>(); + sseRules.add(sseRule); + + ServerSideEncryptionConfiguration sseConf = new ServerSideEncryptionConfiguration(); + sseConf.setRules(sseRules); + + eRequest.setServerSideEncryptionConfiguration(sseConf); + s3client.setBucketEncryption(eRequest); + return true; + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + @Override + public boolean deleteBucketEncryption(BucketTO bucket, long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + s3client.deleteBucketEncryption(bucket.getName()); + return true; + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + @Override + public boolean setBucketVersioning(BucketTO bucket, long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + BucketVersioningConfiguration vConf = new BucketVersioningConfiguration(BucketVersioningConfiguration.ENABLED); + s3client.setBucketVersioningConfiguration( + new SetBucketVersioningConfigurationRequest(bucket.getName(), vConf)); + return true; + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + @Override + public boolean deleteBucketVersioning(BucketTO bucket, long storeId) { + AmazonS3 s3client = getS3ClientByStoreId(storeId); + try { + BucketVersioningConfiguration vConf = new BucketVersioningConfiguration(BucketVersioningConfiguration.SUSPENDED); + s3client.setBucketVersioningConfiguration( + new SetBucketVersioningConfigurationRequest(bucket.getName(), vConf)); + return true; + } catch (AmazonClientException e) { + throw new CloudRuntimeException(e); + } + } + + /** + * Set the bucket quota via the SeaweedFS S3 extension + * ({@code PUT /{bucket}?seaweedfs-quota}), signed with the store admin + * S3 credentials. + * + * SeaweedFS enforces bucket quota server-side by setting a read-only flag + * when usage exceeds the configured limit. The quota is configured via the + * SeaweedFS S3 extension endpoint PUT /{bucket}?seaweedfs-quota, + * authenticated via standard S3 SigV4 and authorized via the + * s3:PutBucketQuota IAM permission. + * + * @param size the GiB size to set the quota to. 0 disables quota. + * @throws CloudRuntimeException if the S3 endpoint or credentials are missing or the request fails. + */ + @Override + public void setBucketQuota(BucketTO bucket, long storeId, long size) { + String s3Url = getS3Url(storeId); + String accessKey = getAccessKey(storeId); + String secretKey = getSecretKey(storeId); + if (s3Url == null || s3Url.isEmpty() || accessKey == null || accessKey.isEmpty() || secretKey == null || secretKey.isEmpty()) { + throw new CloudRuntimeException("SeaweedFS S3 URL and credentials are required to set bucket quota. " + + "Configure 's3Url', 'accesskey', and 'secretkey' in the object store details."); + } + SeaweedFSObjectStoreUtil.setBucketQuotaViaS3Extension(s3Url, accessKey, secretKey, bucket.getName(), size, getS3ExtensionHttpClient()); + } + + /** + * Returns the HTTP client used to send SeaweedFS S3 extension requests + * (e.g. PUT /{bucket}?seaweedfs-quota). Exposed as a protected seam so + * tests can inject a mock client and assert the signed request without + * touching the network. + */ + protected java.net.http.HttpClient getS3ExtensionHttpClient() { + return SeaweedFSObjectStoreUtil.newS3ExtensionHttpClient(); + } + + @Override + public Map<String, Long> getAllBucketsUsage(long storeId) { + Map<String, Long> bucketUsage = new HashMap<>(); + List<BucketVO> bucketList = _bucketDao.listByObjectStoreId(storeId); + if (bucketList.isEmpty()) { + return bucketUsage; + } + + // If the operator has configured a Prometheus metricsUrl, scrape + // per-bucket sizes from the /metrics endpoint in a single HTTP GET. + // This is O(buckets) and avoids the O(total objects) ListObjectsV2 + // scan that doesn't scale to large deployments. Falls back to + // ListObjectsV2 when metricsUrl is not configured. + String metricsUrl = getMetricsUrl(storeId); + if (metricsUrl != null) { + java.util.Set<String> bucketNames = new java.util.HashSet<>(); + for (BucketVO bucket : bucketList) { + bucketNames.add(bucket.getName()); + } + try { + return SeaweedFSObjectStoreUtil.parseBucketUsageFromMetrics( + metricsUrl, bucketNames, getS3ExtensionHttpClient()); Review Comment: The metrics parser intentionally omits managed buckets that are absent or unparseable, but `BucketApiServiceImpl.getBucketUsage()` only updates and sums entries when `size != null`. Returning this partial map therefore leaves stale bucket sizes and under-reports object-store usage instead of applying the documented zero value. Populate every managed bucket with zero before returning (or treat missing samples as a scrape failure so the S3 fallback runs). ########## plugins/storage/object/seaweedfs/src/main/java/org/apache/cloudstack/storage/datastore/util/SeaweedFSObjectStoreUtil.java: ########## @@ -0,0 +1,555 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. +// SPDX-License-Identifier: Apache-2.0 +package org.apache.cloudstack.storage.datastore.util; + +import org.apache.commons.lang3.StringUtils; + +import com.amazonaws.AmazonServiceException; +import com.amazonaws.auth.AWSStaticCredentialsProvider; +import com.amazonaws.auth.BasicAWSCredentials; +import com.amazonaws.client.builder.AwsClientBuilder; +import com.amazonaws.services.identitymanagement.AmazonIdentityManagement; +import com.amazonaws.services.identitymanagement.AmazonIdentityManagementClientBuilder; +import com.amazonaws.services.s3.AmazonS3; +import com.amazonaws.services.s3.AmazonS3ClientBuilder; +import com.cloud.utils.exception.CloudRuntimeException; + +/** + * Utility class for the SeaweedFS object storage provider. + * + * SeaweedFS exposes both an S3-compatible API and an AWS IAM-compatible API, + * so this provider needs no proprietary admin client — only the AWS S3 and IAM + * SDKs, the same pair Cloudian HyperStore already uses in this tree. + */ +public class SeaweedFSObjectStoreUtil { + + /** The name of our Object Store Provider */ + public static final String OBJECT_STORE_PROVIDER_NAME = "SeaweedFS"; + + public static final String STORE_KEY_PROVIDER_NAME = "providerName"; + public static final String STORE_KEY_URL = "url"; + public static final String STORE_KEY_NAME = "name"; + public static final String STORE_KEY_SIZE = "size"; + public static final String STORE_KEY_DETAILS = "details"; + + // Store Details Map key names - managed outside of plugin + public static final String STORE_DETAILS_KEY_ACCESS_KEY = "accesskey"; // admin/root access key + public static final String STORE_DETAILS_KEY_SECRET_KEY = "secretkey"; // admin/root secret key + public static final String STORE_DETAILS_KEY_S3_URL = "s3Url"; // S3 endpoint URL + public static final String STORE_DETAILS_KEY_IAM_URL = "iamUrl"; // IAM endpoint URL + public static final String STORE_DETAILS_KEY_METRICS_URL = "metricsUrl"; // Prometheus metrics endpoint URL (optional, for scalable usage reporting) + + // Account Detail Map key names - credentials created per CloudStack account. + // Namespaced by store ID so one account can use multiple SeaweedFS pools + // without the second pool overwriting the first pool's credentials. + public static final String KEY_ACCESS_KEY_PREFIX = "swfs_AccessKey_"; + public static final String KEY_SECRET_KEY_PREFIX = "swfs_SecretKey_"; + + /** + * Build the account-detail key for the IAM access key of a given store. + */ + public static String keyAccessKey(long storeId) { + return KEY_ACCESS_KEY_PREFIX + storeId; + } + + /** + * Build the account-detail key for the IAM secret key of a given store. + */ + public static String keySecretKey(long storeId) { + return KEY_SECRET_KEY_PREFIX + storeId; + } + + /** + * Connect timeout for the S3 extension HTTP client, in seconds. + */ + public static final int S3_EXTENSION_CONNECT_TIMEOUT_SECONDS = 10; + /** + * Per-request timeout for the S3 extension HTTP request, in seconds. + */ + public static final int S3_EXTENSION_REQUEST_TIMEOUT_SECONDS = 30; + + /** + * IAM user policy name applied to each per-account IAM user. + */ + public static final String IAM_USER_POLICY_NAME = "CloudStackPolicy"; + + /** + * Build an IAM user policy that grants full S3 access only to the given + * buckets (both the bucket and its contents), while denying bucket + * creation and deletion everywhere so CloudStack retains control of the + * bucket lifecycle. When no buckets are provided, all S3 access is denied. + * + * <p>This is the tenant boundary: each account's IAM credentials can only + * operate on that account's own buckets, not on every bucket in the + * SeaweedFS pool. The policy is refreshed whenever buckets are created or + * deleted (see + * {@code SeaweedFSObjectStoreDriverImpl.updateAccountIAMPolicy}). + * + * @param bucketNames the bucket names the account is allowed to access + * @return a JSON IAM policy document + */ + public static String buildAccountIAMPolicy(java.util.List<String> bucketNames) { + StringBuilder sb = new StringBuilder(); + sb.append("{\n"); + sb.append(" \"Version\": \"2012-10-17\",\n"); + sb.append(" \"Statement\": [\n"); + if (bucketNames == null || bucketNames.isEmpty()) { + // No buckets: deny all S3 access. A Resource cannot be empty in + // an IAM policy, so deny everything explicitly. + sb.append(" {\n"); + sb.append(" \"Sid\": \"DenyAllS3\",\n"); + sb.append(" \"Effect\": \"Deny\",\n"); + sb.append(" \"Action\": [\"s3:*\"],\n"); + sb.append(" \"Resource\": [\"arn:aws:s3:::*\", \"arn:aws:s3:::*/*\"]\n"); + sb.append(" }\n"); + } else { + sb.append(" {\n"); + sb.append(" \"Sid\": \"AllowAccountBuckets\",\n"); + sb.append(" \"Effect\": \"Allow\",\n"); + sb.append(" \"Action\": [\"s3:*\"],\n"); + sb.append(" \"Resource\": [\n"); + for (int i = 0; i < bucketNames.size(); i++) { + String name = bucketNames.get(i); + sb.append(" \"arn:aws:s3:::").append(name).append("\",\n"); + sb.append(" \"arn:aws:s3:::").append(name).append("/*\""); + if (i < bucketNames.size() - 1) { + sb.append(","); + } + sb.append("\n"); + } + sb.append(" ]\n"); + sb.append(" }\n"); + } + // Always deny bucket creation/deletion and quota mutation — + // CloudStack controls lifecycle and resource accounting. Denying + // s3:PutBucketQuota prevents a tenant from using the credentials + // returned in BucketResponse to call the SeaweedFS quota extension + // directly and bypass CloudStack's resource accounting. + sb.append(" ,{\n"); + sb.append(" \"Sid\": \"DenyBucketLifecycleAndQuota\",\n"); + sb.append(" \"Effect\": \"Deny\",\n"); + sb.append(" \"Action\": [\"s3:CreateBucket\", \"s3:DeleteBucket\", \"s3:PutBucketQuota\"],\n"); + sb.append(" \"Resource\": \"*\"\n"); + sb.append(" }\n"); + sb.append(" ]\n"); + sb.append("}\n"); + return sb.toString(); + } + + /** + * Returns an S3 connection for the given endpoint and credentials. + * Uses path-style access, which SeaweedFS requires. + * + * @param url the url of the S3 service + * @param accessKey the credentials to use for the S3 connection. + * @param secretKey the matching secret key. + * @return an S3 connection (never null) + * @throws CloudRuntimeException on failure. + */ + public static AmazonS3 getS3Client(String url, String accessKey, String secretKey) { + AmazonS3 client = AmazonS3ClientBuilder.standard() + .enablePathStyleAccess() + .withCredentials(new AWSStaticCredentialsProvider(new BasicAWSCredentials(accessKey, secretKey))) + .withEndpointConfiguration(new AwsClientBuilder.EndpointConfiguration(url, "us-east-1")) + .build(); + if (client == null) { + throw new CloudRuntimeException("Error while creating SeaweedFS S3 client"); + } + return client; + } + + /** + * Returns an IAM connection for the given endpoint and credentials. + * + * @param url the url of the IAM service + * @param accessKey the credentials to use for the iam connection. + * @param secretKey the matching secret key. + * @return an IAM connection (never null) + * @throws CloudRuntimeException on failure. + */ + public static AmazonIdentityManagement getIAMClient(String url, String accessKey, String secretKey) { + AmazonIdentityManagement iamClient = AmazonIdentityManagementClientBuilder.standard() + .withCredentials(new AWSStaticCredentialsProvider(new BasicAWSCredentials(accessKey, secretKey))) + .withEndpointConfiguration(new AwsClientBuilder.EndpointConfiguration(url, "us-east-1")) + .build(); + if (iamClient == null) { + throw new CloudRuntimeException("Error while creating SeaweedFS IAM client"); + } + return iamClient; + } + + /** + * Test the S3Url to confirm it behaves like an S3 Service. + * + * Uses bad credentials and looks for the particular error from S3 that says + * InvalidAccessKeyId was used. Quietly returns if we connect and get the + * expected error back. + * + * @param s3Url the url to check + * @throws CloudRuntimeException if there is any unexpected issue. + */ + public static void validateS3Url(String s3Url) { + try { + AmazonS3 s3Client = SeaweedFSObjectStoreUtil.getS3Client(s3Url, "unknown", "unknown"); + s3Client.listBuckets(); + } catch (AmazonServiceException e) { + if (StringUtils.compareIgnoreCase(e.getErrorCode(), "InvalidAccessKeyId") != 0 + && StringUtils.compareIgnoreCase(e.getErrorCode(), "SignatureDoesNotMatch") != 0) { + throw new CloudRuntimeException("Unexpected response from S3 Endpoint.", e); + } + } + } + + /** + * Test the IAMUrl to confirm it behaves like an IAM Service. + * + * Uses bad credentials and looks for the particular error from IAM that says + * InvalidAccessKeyId or InvalidClientTokenId was used. Quietly returns if we + * connect and get the expected error back. + * + * @param iamUrl the url to check + * @throws CloudRuntimeException if there is any unexpected issue. + */ + public static void validateIAMUrl(String iamUrl) { + try { + AmazonIdentityManagement iamClient = SeaweedFSObjectStoreUtil.getIAMClient(iamUrl, "unknown", "unknown"); + iamClient.listAccessKeys(); + } catch (AmazonServiceException e) { + if (! StringUtils.equalsAnyIgnoreCase(e.getErrorCode(), "InvalidAccessKeyId", "InvalidClientTokenId", "SignatureDoesNotMatch")) { + throw new CloudRuntimeException("Unexpected response from IAM Endpoint.", e); + } + } + } + + /** + * Set bucket quota via the SeaweedFS S3 extension endpoint. + * + * SeaweedFS exposes a custom S3 subresource at + * PUT /{bucket}?seaweedfs-quota + * authenticated via standard S3 SigV4 and authorized via the + * s3:PutBucketQuota IAM permission. This avoids the need for a + * separate admin API credential. + * + * The request body is JSON: + * {"quota_size": <n>, "quota_unit": "GB", "quota_enabled": true} + * + * @param s3Url the S3 endpoint URL (e.g. http://host:8333) + * @param accessKey the S3 access key (must have s3:PutBucketQuota permission) + * @param secretKey the S3 secret key + * @param bucketName the bucket name + * @param sizeGiB the quota size in GiB (0 to disable quota) + * @throws CloudRuntimeException on any failure + */ + public static void setBucketQuotaViaS3Extension(String s3Url, String accessKey, String secretKey, String bucketName, long sizeGiB) { + setBucketQuotaViaS3Extension(s3Url, accessKey, secretKey, bucketName, sizeGiB, newS3ExtensionHttpClient()); + } + + /** + * Build a bounded HTTP client for SeaweedFS S3 extension requests with a + * connect timeout so a stalled endpoint cannot block the management-server + * API thread indefinitely. + */ + public static java.net.http.HttpClient newS3ExtensionHttpClient() { + return java.net.http.HttpClient.newBuilder() + .connectTimeout(java.time.Duration.ofSeconds(S3_EXTENSION_CONNECT_TIMEOUT_SECONDS)) + .build(); + } + + /** + * Set bucket quota via the SeaweedFS S3 extension endpoint using the + * supplied HTTP client. The client is injected so tests can assert the + * signed request without hitting the network. + */ + public static void setBucketQuotaViaS3Extension(String s3Url, String accessKey, String secretKey, + String bucketName, long sizeGiB, java.net.http.HttpClient httpClient) { + if (sizeGiB < 0) { + // Only zero disables a quota; a negative value would corrupt + // resource accounting (BucketApiServiceImpl persists the requested + // value and computes deltas from it), so reject it outright. + throw new CloudRuntimeException("Bucket quota cannot be negative: " + sizeGiB); + } + String body; + if (sizeGiB == 0) { + body = "{\"quota_size\":0,\"quota_unit\":\"B\",\"quota_enabled\":false}"; + } else { + body = String.format("{\"quota_size\":%d,\"quota_unit\":\"GB\",\"quota_enabled\":true}", sizeGiB); + } + try { + executeSignedS3Request("PUT", s3Url, "/" + bucketName + "?seaweedfs-quota", accessKey, secretKey, body, httpClient); + } catch (CloudRuntimeException e) { + // A quota of 0 disables the quota, which is the default state for + // a newly created bucket. If the SeaweedFS quota extension is not + // available (404/405), tolerate the failure for quota 0 so basic + // bucket CRUD works on deployments without the extension. A + // positive quota still requires the extension and must fail. + // + // Distinguish "extension not available" from "bucket not found": + // SeaweedFS returns a standard S3 NoSuchBucket error (with + // <Code>NoSuchBucket</Code> in the body) when the bucket does not + // exist, which must NOT be swallowed — it indicates CloudStack and + // S3 are out of sync. + if (sizeGiB == 0 && e.getMessage() != null + && (e.getMessage().contains("status 404") || e.getMessage().contains("status 405")) + && !e.getMessage().contains("NoSuchBucket")) { + org.apache.logging.log4j.LogManager.getLogger(SeaweedFSObjectStoreUtil.class) + .warn("SeaweedFS quota extension not available for bucket {}; skipping quota disable (quota is already off by default)", bucketName); + return; + } + throw e; + } + } + + /** + * Execute a custom S3 request with SigV4 signing. + * + * Uses the AWS SDK v1 Aws4Signer to sign the request, then sends it via + * java.net.http.HttpClient. This allows calling SeaweedFS-specific S3 + * extensions (like ?seaweedfs-quota) that the AWS SDK doesn't natively + * support. + * + * The query string portion of {@code resourcePath} (e.g. + * {@code /bucket?seaweedfs-quota}) is split off and added to the request + * via {@code addParameter(...)} before signing, so the signer includes it + * in the canonical query string. {@code DefaultRequest.setResourcePath} + * does not parse an embedded query string, so passing it verbatim would + * leave the subresource unsigned while the outgoing URI would still carry + * it, causing a signature mismatch on the server. + * + * @param method HTTP method (PUT, GET, etc.) + * @param s3Url the S3 endpoint base URL + * @param resourcePath the path + optional query string (e.g. /bucket?seaweedfs-quota) + * @param accessKey S3 access key + * @param secretKey S3 secret key + * @param body the request body (null for GET) + * @param httpClient the HTTP client used to send the request + * @return the response body as a string + * @throws CloudRuntimeException on any failure + */ + protected static String executeSignedS3Request(String method, String s3Url, String resourcePath, + String accessKey, String secretKey, String body, + java.net.http.HttpClient httpClient) { + try { + java.net.URI endpointUri = java.net.URI.create(s3Url); + + // Split the resource path into a path and a query string so the + // query parameters are signed as canonical query parameters. + String path = resourcePath; + String queryString = ""; + int q = resourcePath.indexOf('?'); + if (q >= 0) { + path = resourcePath.substring(0, q); + queryString = resourcePath.substring(q + 1); + } + + // Prepend the endpoint's path prefix (e.g. /object-s3) to the + // resource path so the SigV4 canonical URI matches the outgoing + // request URI. Without this, a path-prefixed endpoint behind a + // reverse proxy would sign /bucket but send /object-s3/bucket, + // causing SignatureDoesNotMatch. + String endpointPath = endpointUri.getPath(); + if (endpointPath == null) { + endpointPath = ""; + } + if (endpointPath.endsWith("/")) { + endpointPath = endpointPath.substring(0, endpointPath.length() - 1); + } + String signedResourcePath = endpointPath + path; + + // Build AWS SDK v1 Request for SigV4 signing + com.amazonaws.DefaultRequest<?> request = new com.amazonaws.DefaultRequest<>("s3"); + request.setEndpoint(endpointUri); + request.setHttpMethod(com.amazonaws.http.HttpMethodName.valueOf(method)); + request.setResourcePath(signedResourcePath); + if (! queryString.isEmpty()) { + for (String pair : queryString.split("&")) { + if (pair.isEmpty()) { + continue; + } + int eq = pair.indexOf('='); + if (eq >= 0) { + request.addParameter(pair.substring(0, eq), pair.substring(eq + 1)); + } else { + request.addParameter(pair, ""); + } + } + } + if (body != null) { + byte[] bodyBytes = body.getBytes(java.nio.charset.StandardCharsets.UTF_8); + request.setContent(new java.io.ByteArrayInputStream(bodyBytes)); + request.getHeaders().put("Content-Length", String.valueOf(bodyBytes.length)); + request.getHeaders().put("Content-Type", "application/json"); + } + + // Sign with SigV4 (AWSS3V4Signer, not the legacy S3Signer which is SigV2) + com.amazonaws.auth.AWSCredentials credentials = new com.amazonaws.auth.BasicAWSCredentials(accessKey, secretKey); + com.amazonaws.services.s3.internal.AWSS3V4Signer signer = new com.amazonaws.services.s3.internal.AWSS3V4Signer(); + signer.setServiceName("s3"); + signer.setRegionName("us-east-1"); + signer.sign(request, credentials); + + // Build and send the HTTP request with signed headers. The URI + // carries the original query string; the signed headers (including + // Authorization) are copied from the signed request. Restricted + // headers (e.g. Content-Length, Host) are set by the HTTP client / + // URI itself and cannot be added via HttpRequest.Builder.header(), + // so they are skipped here. + // Build the outgoing URI preserving the endpoint path prefix (e.g. + // https://host/object-s3) by concatenating it with the resource + // path. This matches the signed resource path so SigV4 verifies. + java.net.URI fullUri = java.net.URI.create( + endpointUri.getScheme() + "://" + endpointUri.getRawAuthority() + + endpointPath + path); + if (! queryString.isEmpty()) { + fullUri = java.net.URI.create(fullUri.toString() + "?" + queryString); + } + java.net.http.HttpRequest.Builder reqBuilder = java.net.http.HttpRequest.newBuilder() + .uri(fullUri) + .timeout(java.time.Duration.ofSeconds(S3_EXTENSION_REQUEST_TIMEOUT_SECONDS)); + for (java.util.Map.Entry<String, String> entry : request.getHeaders().entrySet()) { + String headerName = entry.getKey(); + if (headerName == null || entry.getValue() == null) { + continue; + } + if (isRestrictedHttpHeader(headerName)) { + continue; + } + reqBuilder.header(headerName, entry.getValue()); + } + if (body != null) { + reqBuilder.method(method, java.net.http.HttpRequest.BodyPublishers.ofString(body)); + } else { + reqBuilder.method(method, java.net.http.HttpRequest.BodyPublishers.noBody()); + } + + java.net.http.HttpResponse<String> response = httpClient.send(reqBuilder.build(), + java.net.http.HttpResponse.BodyHandlers.ofString()); + + int statusCode = response.statusCode(); + if (statusCode < 200 || statusCode >= 300) { + throw new CloudRuntimeException(String.format( + "S3 extension request %s %s failed with status %d: %s", + method, fullUri, statusCode, response.body())); + } + return response.body(); + } catch (CloudRuntimeException e) { + throw e; + } catch (Exception e) { + throw new CloudRuntimeException("S3 extension request failed: " + method + " " + resourcePath, e); + } + } + + /** + * Headers that {@code java.net.http.HttpRequest.Builder.header()} rejects + * because they are managed by the HTTP client itself (content length is + * derived from the body publisher, host from the URI, etc.). They must be + * skipped when copying the signed headers onto the outgoing request. + */ + private static boolean isRestrictedHttpHeader(String headerName) { + if (headerName == null) { + return true; + } + switch (headerName.toLowerCase(java.util.Locale.ROOT)) { + case "content-length": + case "host": + case "connection": + case "expect": + case "upgrade": + return true; + default: + return false; + } + } + + /** + * Prometheus metric name for per-bucket logical size. SeaweedFS publishes + * this gauge from the S3 API server's bucket-size metrics loop. + */ + public static final String METRIC_BUCKET_SIZE_BYTES = "seaweed_s3_bucket_size_bytes"; + + /** + * Scrape the SeaweedFS Prometheus {@code /metrics} endpoint and return a + * map of bucket name to logical size in bytes. + * + * <p>This is a single HTTP GET that returns all bucket sizes in O(buckets) + * time, replacing the O(total objects) {@code ListObjectsV2} scan used as a + * fallback. The operator configures {@code metricsUrl} as a store detail + * pointing at the SeaweedFS S3 server's metrics port (or a Prometheus + * server that scrapes it). Review Comment: `metricsUrl` is documented as accepting a Prometheus server, but the implementation appends `/metrics` and parses the response as SeaweedFS exporter text. A Prometheus server's `/metrics` endpoint exposes Prometheus's own metrics, not the scraped `seaweed_s3_bucket_size_bytes` series, so this configuration produces an empty/partial map and suppresses the S3 fallback. Query the Prometheus HTTP API instead, or restrict the contract to a SeaweedFS exporter endpoint. -- 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]
