Oracle International Corporation

Jonathan Giloni

Ravi Shankar Thammaiah

Kevin S Neel

Carol L Colrain

Douglas N Surber

Tong Zhou

Arturo Ismael Quinto

Raunak Rungta

Juan R Loaiza

Yunrui Li

Karl Dias

Sumanta Kumar Chatterjee

Lakshminarayanan Luxi Chidambaran

Srinath Krishnaswamy

Leonidas Galanis

Mehul Dilip Bastawala

Stephen John Vivian

Yujun Wang

Krishna Mohan Itikarlapalli

Romain Colle

Efstratios Papadomanolakis

Sreekumar K Seshadri

Supiti Buranawatanachoke

Xiaoli Qi

William Habeck

Manoj Dhanapal

Michael Cusson

Scot McKinley

Techniques are described for preserving the inflight sessions failing over from a primary database to the replicated logical database of the primary database. In an implementation, prior to failover, when the primary database server receives a commit for a transaction, the process stores a commit indication that the transaction has been committed by performing a corresponding SQL command. The commit indication is replicated to the logical replica database by virtue of the replication of the SQL command and its execution on the logical replica database. Accordingly, the standby database server in the failover session may successfully request for the outcome of the transaction. Techniques are also described for the client-side LOB references to be preserved when failing over to the logical replica database, for AS OF queries preserved, and for versioning of checksums, signatures and structures across logical replicas.

Failover of database sessions to a logical replica database

Techniques are disclosed herein for identifying, recording and restoring the state of a database session and various aspects thereof. A session template data structure is generated that includes session attribute values describing various aspects of the session that is established between a client system and a database management system (DBMS and enables the client system to issue to the DBMS commands for execution. Based on the session attribute values, DBMS may generate a template identifier corresponding to the session template data structure. The template identifier may be stored in an association with the session state that it partially (or in whole) represents. In an embodiment, when another state of a session is captured, if the template identifier for the state is the same, then rather than storing the attribute-value pairs for the other state, the template identifier is further associated with the other state. In an embodiment, a request boundary is detected where the session is known to be at a recoverable point. If recovery of the session is needed, the session state is restored, and replay of commands start from this point. Each command replayed is verified to produce the same session state as it produced by the original execution. If the session is determined to be a safe point, then all the commands recorded for replay prior to the safe point may be deleted. In an embodiment, the template is used to set the initial state when borrowing from a session pool. The state tracking is also used to identify that the session can be failed over safely during planned operation as the session is unlikely to drain by itself even when not used.

Session state tracking

Techniques are disclosed herein for identifying, recording and restoring the state of a database session and various aspects thereof. A session template data structure is generated that includes session attribute values describing various aspects of the session that is established between a client system and a database management system (DBMS and enables the client system to issue to the DBMS commands for execution. Based on the session attribute values, DBMS may generate a template identifier corresponding to the session template data structure. The template identifier may be stored in an association with the session state that it partially (or in whole) represents. In an embodiment, when another state of a session is captured, if the template identifier for the state is the same, then rather than storing the attribute-value pairs for the other state, the template identifier is further associated with the other state. In an embodiment, a request boundary is detected where the session is known to be at a recoverable point. If recovery of the session is needed, the session state is restored, and replay of commands start from this point. Each command replayed is verified to produce the same session state as it produced at original execution. If the session is determined to be a safe point, then all the commands recorded for replay prior to the safe point may be deleted. In an embodiment, the template is used to set the initial state when borrowing from a session pool The state tracking is also used to know that the session can be failed over safely during planned operation as the session is unlikely to drain by itself even when not used.

Session templates

Described is an improved approach to implement parallel queries where session states are saved for parallelization resources. When work needs to be performed in the parallel query system for a given session, a search can be performed to identify a resource (from among the pool of available resources) that had previously been used by that session, and which had saved a session state object for that previous connection to the session. Instead of incurring the entirety of setup costs each time workload is assigned to a resource, the saved session state can be used to re-set the context for the resource to the configuration requirements for that session.

Method and system for implementing parallel database queries

Systems for managing shared computing resources. In a multi-process computing environment a concurrency object data structure pertaining to a shared resource is made available to be accessed by two or more processing entities. The concurrency object comprises a consecutive read count that tracks the number of consecutive read requests that have been received for shared read access to the shared resource. A shared concurrency access state is entered based on comparison of the consecutive read count to a threshold value. Entering the shared concurrency access state begins a period during which grant of further shared access requests do not require semaphore operations or other atomic operations that pertains to the shared resource.

Lockless execution in read-mostly workloads for efficient concurrent process execution on shared resources

A method for memory quarantine comprises: in response to a memory error, determining an extent associated with the memory error, wherein the extent defines an amount of memory comprising a plurality of chunks; traversing, forward from a beginning chunk of the plurality chunks, the plurality of chunks until a first chunk is determined to be corrupt; traversing, in reverse from a last chunk of the plurality of chunks, the plurality of chunks until a second chunk is determine to be corrupt; flagging a subset of the plurality of chunks as quarantined, wherein the subset begins with the first chunk and ends with the second chunk, and wherein quarantine prevents the subset from being reallocated after being released.

Memory quarantine

A server side connection pool provides pooling of server connections. Thus, client side processes do not create groups of dedicated resources. Clients are provided the capability to communicate to a server side connection broker process to GET a pooled server connection comprising connection state plus a thread of execution for doing some work, whereby requesting the work can bypass the connection broker process and go directly to the assigned pooled connection. Once the client issues a RELEASE to release the pooled connection back to the pool, a completely different client process on potentially a different client middle tier machine can reuse the same pooled connection. The server-side connection pool can be logically partitioned into sub-pools that correspond to connection classes that typically correspond to respective applications. A client process can request a pooled connection having a particular state, where a stored procedure can be executed to generate such state.

Server-side connection resource pooling

A method, apparatus, and computer-readable medium is provided for replaying captured workload data in a database environment with connection pools. A replay system includes a dispatcher that manages a connection pool. In one aspect, the dispatcher selects available connections for sending database commands that are ready to be executed against a database, but prevents from being sent, or holds, commands that are not ready to be executed against the database. Commands may be stored at the dispatcher until the database advances to a state that allows the commands to be executed. In another aspect, the dispatcher may use an available connection to send a command to a database instance even though the command is not ready to be executed, as long as at least a threshold number or percentage of connections remains available.

Transactionally consistent database replay in an environment with connection pooling

Growing community of inventors

Brooklyn, NY, United States of America

Jonathan Giloni