' & $ Chapter 15: Recovery System · Failure Classification · Storage Structure · Recovery and Atomicity · Log-Based Recovery · Shadow Paging · Recovery With Concurrent Transactions · Buffer Management · Failure with Loss of Nonvolatile Storage · Advanced Recovery Techniques Database Systems Concepts 15.1 Silberschatz, Korth and Sudarshan c 1997 ' & $ Failure Classification · Transaction failure : ­ Logical errors: transaction cannot complete due to some internal error condition ­ System errors: the database system must terminate an active transaction due to an error condition (e.g., deadlock) · System crash: a power failure or other hardware or software failure causes the system to crash. It is assumed that non-volatile storage contents are not corrupted. · Disk failure: a head crash or similar failure destroys all or part of disk storage Database Systems Concepts 15.2 Silberschatz, Korth and Sudarshan c 1997 ' & $ Storage Structure · Volatile storage: ­ does not survive system crashes ­ examples: main memory, cache memory · Nonvolatile storage: ­ survives system crashes ­ examples: disk, tape · Stable storage: ­ a mythical form of storage that survives all failures ­ approximated by maintaining multiple copies on distinct nonvolatile media Database Systems Concepts 15.3 Silberschatz, Korth and Sudarshan c 1997 ' & $ Stable-Storage Implementation · Maintain multiple copies of each block on separate disks; copies can be at remote sites to protect against disasters such as fire or flooding. · Failure during data transfer can result in inconsistent copies · Protecting storage media from failure during data transfer (one solution): ­ Execute output operation as follows (assuming two copies of each block): 1. Write the information onto the first physical block. 2. When the first write successfully completes, write the same information onto the second physical block. 3. The output is completed only after the second write successfully completes. Database Systems Concepts 15.4 Silberschatz, Korth and Sudarshan c 1997 ' & $ Stable-Storage Implementation (Cont.) · Protecting storage media from failure during data transfer (cont.): ­ Copies of a block may differ due to failure during output operation. To recover from failure: 1. First find inconsistent blocks: (a) Expensive solution: Compare the two copies of every disk block. (b) Better solution: Record in-progress disk writes on non-volatile storage. Use this information during recovery to find blocks that may be inconsistent, and only compare copies of these. 2. If either copy of an inconsistent block is detected to have an error (bad checksum), overwrite it by the other copy. If both have no error, but are different, overwrite the second block by the first block. Database Systems Concepts 15.5 Silberschatz, Korth and Sudarshan c 1997 ' & $ Data Access · Physical blocks are those blocks residing on the disk. Buffer blocks are the blocks residing temporarily in main memory. · Block movements between disk and main memory are initiated through the following two operations: ­ input(B) transfers the physical block B to main memory. ­ output(B) transfers the buffer block B to the disk, and replaces the appropriate physical block there. · Each transaction Ti has its private work-area in which local copies of all data items accessed and updated by it are kept. Ti 's local copy of a data item X is called xi. · We assume, for simplicity, that each data item fits in, and is stored inside, a single block. Database Systems Concepts 15.6 Silberschatz, Korth and Sudarshan c 1997 ' & $ Data Access (Cont.) · Transaction transfers data items between system buffer blocks and its private work-area using the following operations : ­ read(X) assigns the value of data item X to the local variable xi. ­ write(X) assigns the value of local variable xi to data item X in the buffer block. ­ both these commands may necessitate the issue of an input(BX ) instruction before the assignment, if the block BX in which X resides is not already in memory. · Transactions perform read(X) while accessing X for the first time; all subsequent accesses are to the local copy. After last access, transaction executes write(X). · output(BX ) need not immediately follow write(X). System can perform the output operation when it deems fit. Database Systems Concepts 15.7 Silberschatz, Korth and Sudarshan c 1997 ' & $ Example of Data Access T1 T2 write(B) read(A) write(B) read(A) global buffer local bufferlocal buffer AA B memory disk actual A B A B Database Systems Concepts 15.8 Silberschatz, Korth and Sudarshan c 1997 ' & $ Recovery and Atomicity · Consider transaction Ti that transfers $50 from account A to account B; goal is either to perform all database modifications made by Ti or none at all. · Several output operations may be required for Ti (to output A and B). A failure may occur after one of these modifications have been made but before all of them are made. · To ensure atomicity despite failures, we first output information describing the modifications to stable storage without modifying the database itself. · We study two approaches: ­ log-based recovery, and ­ shadow-paging · We assume (initially) that transactions run serially, that is, one after the other. Database Systems Concepts 15.9 Silberschatz, Korth and Sudarshan c 1997 ' & $ Log-Based Recovery · A log is kept on stable storage. The log is a sequence of log records, and maintains a record of update activities on the database. · When transaction Ti starts, it registers itself by writing a < Ti start > log record · Before Ti executes write(X), a log record < Ti , X, V1, V2 > is written, where V1 is the value of X before the write, and V2 is the value to be written to X. · When Ti finishes it last statement, the log record < Ti commit > is written. · We assume for now that log records are written directly to stable storage (that is, they are not buffered) Database Systems Concepts 15.10 Silberschatz, Korth and Sudarshan c 1997 ' & $ Deferred Database Modification · This scheme ensures atomicity despite failures by recording all modifications to log, but deferring all the writes to after partial commit. · Assume that transactions execute serially · Transaction starts by writing < Ti start > record to log. · A write(X) operation results in a log record < Ti , X, V > being written, where V is the new value for X. The write is not performed on X at this time, but is deferred. · When Ti partially commits, < Ti commit > is written to the log · Finally, log records are used to actually execute the previously deferred writes. Database Systems Concepts 15.11 Silberschatz, Korth and Sudarshan c 1997 ' & $ Deferred Database Modification (Cont.) · During recovery after a crash, a transaction needs to be redone if and only if both < Ti start > and < Ti commit > are there in the log. · Redoing a transaction Ti (redo(Ti )) sets the value of all data items updated by the transaction to the new values. · Crashes can occur while the transaction is executing the original updates, or while recovery action is being taken · example transactions T0 and T1 (T0 executes before T1): T0: read(A) A := A - 50 write(A) read(B) B := B + 50 write(B) T1: read(C) C := C - 100 write(C) Database Systems Concepts 15.12 Silberschatz, Korth and Sudarshan c 1997 ' & $ Deferred Database Modification (Cont.) · Below we show the log as it appears at three instances of time. (a) (b) (c) · If log on stable storage at time of crash is as in case: (a) No redo actions need to be taken (b) redo(T0) must be performed since is present (c) redo(T0) must be performed followed by redo(T1) since and are present Database Systems Concepts 15.13 Silberschatz, Korth and Sudarshan c 1997 ' & $ Immediate Database Modification · This scheme allows database updates of an uncommitted transaction to be made as the writes are issued; since undoing may be needed, update logs must have both old value and new value · Update log record must be written before database item is written · Output of updated blocks can take place at any time before or after transaction commit · Order in which blocks are output can be different from the order in which they are written. Database Systems Concepts 15.14 Silberschatz, Korth and Sudarshan c 1997 ' & $ Immediate Database Modification Example Log Write Output A = 950 B = 2050 C = 600 BB, BC BA · Note: BX denotes block containing X. Database Systems Concepts 15.15 Silberschatz, Korth and Sudarshan c 1997 ' & $ Immediate Database Modification (Cont.) · Recovery procedure has two operations instead of one : ­ undo(Ti ) restores the value of all data items updated by Ti to their old values, going backwards from the last log record for Ti ­ redo(Ti ) sets the value of all data items updated by Ti to the new values, going forward from the first log record for Ti · When recovering after failure: ­ Transaction Ti needs to be undone if the log contains the record < Ti start>, but does not contain the record . ­ Transaction Ti needs to be redone if the log contains both the record < Ti start> and the record . · Undo operations are performed first, then redo operations. Database Systems Concepts 15.16 Silberschatz, Korth and Sudarshan c 1997 ' & $ Immediate DB Modification Recovery Example Below we show the log as it appears at three instances of time. (a) (b) (c) Recovery actions in each case above are: (a) undo(T0): B is restored to 2000 and A to 1000. (b) undo(T1) and redo(T0): C is restored to 700, and then A and B are set to 950 and 2050 respectively. (c) redo(T0) and redo(T1): A and B are set to 950 and 2050 respectively. Then C is set to 600. Database Systems Concepts 15.17 Silberschatz, Korth and Sudarshan c 1997 ' & $ Checkpoints · Problems in recovery procedure as discussed earlier : 1. searching the entire log is time-consuming 2. we might unnecessarily redo transactions which have already output their updates to the database. · Streamline recovery procedure by periodically performing checkpointing 1. Output all log records currently residing in main memory onto stable storage. 2. Output all modified buffer blocks to the disk. 3. Write a log record onto stable storage. Database Systems Concepts 15.18 Silberschatz, Korth and Sudarshan c 1997 ' & $ Checkpoints (Cont.) · During recovery we need to consider only the most recent transaction Ti that started before the checkpoint, and transactions that started after Ti . · Scan backwards from end of log to find the most recent record · Continue scanning backwards till a record is found. · Need only consider the part of log following above start record. Earlier part of log can be ignored during recovery, and can be erased whenever desired. · For all transactions (starting from Ti or later) with no , execute undo(Ti ). (Done only in case of immediate modification.) · Scanning forward in the log, for all transactions starting from Ti or later with a , execute redo(Ti ). Database Systems Concepts 15.19 Silberschatz, Korth and Sudarshan c 1997 ' & $ Example of Checkpoints Tc T1 T2 T3 T4 Tf time system failurecheckpoint · T1 can be ignored (updates already output to disk due to checkpoint) · T2 and T3 redone · T4 undone Database Systems Concepts 15.20 Silberschatz, Korth and Sudarshan c 1997 ' & $ Shadow Paging · Alternative to log-based recovery · Idea: maintain two page tables during the lifetime of a trans- action ­ the current page table, and the shadow page table · Store the shadow page table in nonvolatile storage, such that state of the database prior to transaction execution may be recovered. Shadow page table is never modified during execution · To start with, both the page tables are identical. Only current page table is used for data item accesses during execution of the transaction. · Whenever any page is about to be written for the first time, a copy of this page is made onto an unused page. The current page table is then made to point to the copy, and the update is performed on the copy Database Systems Concepts 15.21 Silberschatz, Korth and Sudarshan c 1997 ' & $ Example of Shadow Paging 1 2 3 4 5 6 7 8 9 10 shadow page table current page table 1 2 3 4 5 6 7 8 9 10 pages on disk Shadow and current page tables after write to page 4 Database Systems Concepts 15.22 Silberschatz, Korth and Sudarshan c 1997 ' & $ Shadow Paging (Cont.) · To commit a transaction: 1. Flush all modified pages in main memory to disk 2. Output current page table to disk 3. Make the current page the new shadow page table ­ keep a pointer to the shadow page table at a fixed (known) location on disk. ­ to make the current page table the new shadow page table, simply update the pointer to point to current page table on disk · Once pointer to shadow page table has been written, transaction is committed. · No recovery is needed after a crash -- new transactions can start right away, using the shadow page table. · Pages not pointed to from current/shadow page table should be freed (garbage collected). Database Systems Concepts 15.23 Silberschatz, Korth and Sudarshan c 1997 ' & $ Shadow Paging (Cont.) · Advantages of shadow-paging over log-based schemes ­ no overhead of writing log records; recovery is trivial · Disadvantages : ­ Commit overhead is high (many pages need to be flushed) ­ Data gets fragmented (related pages get separated) ­ After every transaction completion, the database pages containing old versions of modified data need to be garbage collected and put into the list of unused pages ­ Hard to extend algorithm to allow transactions to run concurrently Database Systems Concepts 15.24 Silberschatz, Korth and Sudarshan c 1997 ' & $ Recovery With Concurrent Transactions · We modify the log-based recovery schemes to allow multiple transactions to execute concurrently. ­ All transactions share a single disk buffer and a single log ­ A buffer block can have data items updated by one or more transactions · We assume concurrency control using strict two-phase locking; ie. the updates of uncommitted transactions should not be visible to other transactions · Logging is done as described earlier. Log records of different transactions may be interspersed in the log. · The checkpointing technique and actions taken on recovery have to be changed, since several transactions may be active when a checkpoint is performed. Database Systems Concepts 15.25 Silberschatz, Korth and Sudarshan c 1997 ' & $ Recovery With Concurrent Transactions (Cont.) · Checkpoints are performed as before, except that the checkpoint log record is now of the form , where L is the list of transactions active at the time of the checkpoint. · When the system recovers from a crash, it first does the following: 1. Initialize undo-list and redo-list to empty 2. Scan the log backwards from the end, stopping when the first record is found. For each record found during the scan: ­ if the record is , add Ti to redo-list ­ if the record is , then if Ti is not in redo-list, add Ti to undo-list 3. For every Ti in L, if Ti is not in redo-list, add Ti to undo-list Database Systems Concepts 15.26 Silberschatz, Korth and Sudarshan c 1997 ' & $ Recovery With Concurrent Transactions (Cont.) · At this point undo-list consists of incomplete transactions which must be undone, and redo-list consists of finished transactions that must be redone. · Recovery now continues as follows: 4. Scan log backwards from most recent record, stopping when < Ti start > records have been encountered for every Ti in undo-list. During the scan, perform undo for each log record that belongs to a transaction in undo-list. 5. Locate the most recent record. 6. Scan log forwards from the record till the end of the log. During the scan, perform redo for each log record that belongs to a transaction on redo-list. Database Systems Concepts 15.27 Silberschatz, Korth and Sudarshan c 1997 ' & $ Example of Recovery Go over the steps of the recovery algorithm on the following log: < T0 start> < T1 start> /* Scan in Step 4 stops here*/ < T2 start> < T3 start> Database Systems Concepts 15.28 Silberschatz, Korth and Sudarshan c 1997 ' & $ Log Record Buffering · Log record buffering: log records are buffered in main memory, instead of of being output directly to stable storage. Log records are output to stable storage when a block of log records in the buffer is full, or a log force operation is executed. · Several log records can thus be output using a single output operation, reducing the I/O cost. Database Systems Concepts 15.29 Silberschatz, Korth and Sudarshan c 1997 ' & $ Log Record Buffering (Cont.) · The rules below must be followed if log records are buffered: ­ Log records are output to stable storage in the order in which they are created. ­ Transaction Ti enters the commit state after the log record has been output to stable storage. ­ Before a block of data in main memory is output to the database, all log records pertaining to data in that block must have been output to stable storage. (This rule is called the write-ahead logging or WAL rule.) · As a result of the write-ahead logging rule, if a block with uncommitted updates is output to disk, log records with undo information for the updates are output to the log on stable storage first. Database Systems Concepts 15.30 Silberschatz, Korth and Sudarshan c 1997 ' & $ Buffer Management (Cont.) · Log force is performed to commit a transaction by forcing all its log records (including the commit record) to stable storage. · Our checkpointing algorithm requires that no updates should be in progress on a block when it is output to disk. Can be ensured as follows. ­ Before writing a data item, transaction acquires exclusive lock on block containing the data item ­ Lock can be released once the write is completed. (Such locks held for short duration are called latches.) ­ Before a block is output to disk, the system acquires an exclusive lock on the block Database Systems Concepts 15.31 Silberschatz, Korth and Sudarshan c 1997 ' & $ Buffer Management (Cont.) · Database buffer can be implemented either ­ in an area of real main-memory reserved for the database, or ­ in virtual memory · Implementing buffer in reserved main-memory has drawbacks: ­ Memory is partitioned before-hand between database buffer and applications, limiting flexibility. ­ Needs may change, and although operating system knows best how memory should be divided up at any time, it cannot change the partitioning of memory. Database Systems Concepts 15.32 Silberschatz, Korth and Sudarshan c 1997 ' & $ Buffer Management (Cont.) Database buffers are generally implemented in virtual memory in spite of some drawbacks: · When operating system needs to evict a page that has been modified, the page is written to swap space on disk. · When database decides to write buffer page to disk, buffer page may be in swap space, and may have to be read from disk and written to another location on disk, resulting in extra I/O! (Known as dual paging problem.) · Ideally when swapping out a database buffer page, operating system should pass control to database, which in turn outputs page to database instead of to swap space (making sure to output log records first) ­ Dual paging can thus be avoided, but common operating systems do not support such functionality. Database Systems Concepts 15.33 Silberschatz, Korth and Sudarshan c 1997 ' & $ Failure with Loss of Nonvolatile Storage · Periodically dump the entire content of the database to stable storage · No transaction may be active during the dump procedure; a procedure similar to checkpointing must take place ­ Output all log records currently residing in main memory onto stable storage. ­ Output all buffer blocks onto the disk. ­ Copy the contents of the database to stable storage. ­ Output a record to log on stable storage. · To recover from disk failure, restore database from most recent dump. Then log is consulted and all transactions that committed since the dump are redone. · Can be extended to allow transactions to be active during dump; known as fuzzy or online dump. Database Systems Concepts 15.34 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advanced Recovery Techniques · Support high-concurrency locking techniques, such as those used for B+ -tree concurrency control · Operations like B+ -tree insertions and deletions release locks early. They cannot be undone by restoring old values (physical undo), since once a lock is released, other transactions may have updated the B+ -tree. · Instead, insertions (resp. deletions) are undone by executing a deletion (resp. insertion) operation (known as logical undo). · For such operations, undo log records should contain the undo operation to be executed; called logical undo logging, in contrast to physical undo logging. · Redo information is logged physically (that is, new value for each write) even for such operations. Database Systems Concepts 15.35 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advanced Recovery Techniques (Cont.) Operation logging is done as follows: · When operation starts, log < Ti , Oj , operation-begin>. Here Oj is a unique identifier of the operation instance. · While operation is executing, normal log records with physical redo and physical undo information are logged. · When operation completes, is logged, where U contains information needed to perform a logical undo information. · If crash/rollback occurs before operation completes, the operation-end log record is not found, and the physical undo information is used to undo operation. · If crash/rollback occurs after the operation completes, the operation-end log record is found, and logical undo is performed using U; the physical undo information for the operation is ignored. Database Systems Concepts 15.36 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advanced Recovery Techniques (Cont.) Rollback of transaction Ti is done as follows: · Scan the log backwards 1. If a log record is found, perform the undo and log a special redo-only record < Ti , X, V1 >. 2. If a < Ti , Oj , operation-end, U > record is found ­ Rollback the operation logically using the undo information U. Updates performed during roll back are logged just like during normal operation execution. ­ At the end of the operation rollback, instead of logging an operation-end record, generate a record < Ti , Oj, operation-abort>. ­ Skip all preceding log records for Ti until the record < Ti , Oj, operation-begin> is found Database Systems Concepts 15.37 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advanced Recovery Techniques (Cont.) · Scan the log backwards (cont.): 3. If a redo-only record is found ignore it 4. If a < Ti , Oj, operation-abort> record is found, skip all preceding log records for Ti until the record < Ti , Oj, operation-begin> is found. · Stop the scan when the record < Ti , start> is found · Add a < Ti , abort> record to the log Some points to note: · Cases 3 and 4 above can occur only if the database crashes while a transaction is being rolled back. · Skipping of log records as in case 4 is important to prevent multiple rollback of the same operation. Database Systems Concepts 15.38 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advanced Recovery Techniques (Cont.) The following actions are taken when recovering from system crash 1. Repeat history by physically redoing all updates of all transactions, scanning log forward from last < checkpoint L> record · undo-list is set to L initially · Whenever is found Ti is added to undo-list · Whenever or is found, Ti is deleted from undo-list This brings database to state as of crash, with committed as well as uncommitted transactions having been redone. Now undo-list contains transactions that are incomplete, that is, have neither committed nor been fully rolled back. Database Systems Concepts 15.39 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advanced Recovery Techniques (Cont.) 2. Scan log backwards, performing undo on log records of transactions found in undo-list. Transactions are rolled back as described earlier. · When is found for a transaction Ti in undo-list, write a log record. · Stop scan when records have been found for all Ti in undo-list This undoes the effects of incomplete transactions (those with neither commit nor abort log records). Recovery is now complete. Database Systems Concepts 15.40 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advanced Recovery Techniques (Cont.) · Checkpointing is done as follows: 1. Output all log records in memory to stable storage 2. Output to disk all modified buffer blocks 3. Output to log on stable storage a record. Transactions are not allowed to perform any actions while checkpointing is in progress. · Fuzzy checkpointing allows transactions to progress while the most time consuming parts of checkpointing are in progress Database Systems Concepts 15.41 Silberschatz, Korth and Sudarshan c 1997 ' & $ Advanced Recovery Techniques (Cont.) · Fuzzy checkpointing is done as follows: 1. Write a log record and force log to stable storage 2. Note list M of modified buffer blocks 3. Now permit transactions to proceed with their actions 4. Output to disk all modified buffer blocks in list M ­ blocks should not be updated while being output, and ­ all log records pertaining to a block must be output before the block is output 5. Store a pointer to the checkpoint record in a fixed position last checkpoint on disk · When recovering using a fuzzy checkpoint, start scan from the checkpoint record pointed to by last checkpoint. · Log records before last checkpoint have their updates reflected in database on disk, and need not be redone. Database Systems Concepts 15.42 Silberschatz, Korth and Sudarshan c 1997