Time proposals in director-based database system for transactional consistency
Techniques are disclosed relating to a database system includes worker nodes operable to perform transactions and director nodes operable to ensure transactional consistency for the transactions. A worker node may receive a request to perform a transaction involving writing a record. The worker node may then issue, to director nodes of the database system, a request for information that facilitates performance of an operation for the transaction. A director node may determine whether to approve the request based on whether the operation could cause transactional inconsistency in the database system. The worker node may proceed to perform the operation for the transaction in response to receiving approval responses from a majority of the director nodes, with none of the received responses indicating a disapproval of the transaction.
1. A method, comprising:
receiving, by a worker node of a database system, a request to perform a transaction that involves writing a record, wherein the database system includes a plurality of worker nodes operable to perform transactions for the database system and a plurality of director nodes operable to ensure transactional consistency for the transactions;
selecting, by the worker node, a proposed commit time for the transaction; and
issuing, by the worker node to director nodes of the plurality of director nodes, a request for approval to commit the transaction, wherein the request specifies the proposed commit time, and wherein a given director node of the director nodes is operable to process the request for approval upon reaching the proposed commit time according to a clock of the given director node that identifies a time observed by the given director node.
2. The method of claim 1 , further comprising:
determining, by the worker node, time delays in communicating with the director nodes, wherein a given time delay of the time delays is a delay between sending a given request from the worker node to a director node of the director nodes and the director node receiving the given request, and wherein the proposed commit time is selected based on the determined time delays.
3. The method of claim 2 , further comprising:
recording a first time at which the given request is sent to the director node; and
receiving a response to the given request that identifies a second time at which the given request was received at the director node, wherein the given time delay is determined based on the first and second times and a clock-skew.
4. The method of claim 2 , wherein the proposed commit time is a future time selected, based on the time delays, to allow the request for approval to arrive at the director nodes before the future time according to respective clocks of the director nodes.
5. The method of claim 1 , further comprising:
receiving, by the worker node from at least one of the director nodes, a disapproval response that indicates a disapproval of the proposed commit time based on an arrival of the request at the at least one director node after the proposed commit time according to a clock of the at least one director node.
6. The method of claim 5 , wherein the proposed commit time is a first amount of time after a current time that is identified by a clock of the worker node, and wherein the method further comprises:
based on receiving the disapproval response, the worker node selecting another proposed commit time that is a second amount of time after the current time identified by the clock of the worker node, wherein the second amount of time is greater than the first amount of time.
7. The method of claim 1 , wherein the proposed commit time is selected based on time identified by a clock of the worker node, and wherein the clock of the worker node identifies a different time than a clock of at least one of the plurality of director nodes.
8. The method of claim 1 , wherein a clock of a particular one of the plurality of director nodes specifies a different time than a clock of at least one other director node of the plurality of director nodes.
9. The method of claim 1 , wherein the given director node is operable to check for conflicts between the record of the transaction and a set of approved records known to the given director node that occurred before the proposed commit time.
10. A non-transitory computer readable medium having program instructions stored thereon that are capable of causing a worker node of a database system to perform operations comprising:
receiving a request to perform a transaction that involves writing a record, wherein the database system includes a plurality of worker nodes operable to perform transactions for the database system and a plurality of director nodes operable to ensure transactional consistency for the transactions;
selecting a proposed commit time for the transaction; and
issuing, to director nodes of the plurality of director nodes, a request for approval to commit the transaction, wherein the request specifies the proposed commit time, and wherein a given director node of the director nodes is operable to process the request for approval upon reaching the proposed commit time according to a clock of the given director node that identifies a time observed by the given director node.
11. The medium of claim 10 , wherein the operations further comprise:
maintaining delay information that identifies time delays between sending a given request to the director nodes and the director nodes receiving the given request, wherein the selecting of the proposed commit time is based on the time delays.
12. The medium of claim 11 , wherein the operations further comprise:
recording a first time at which the request for approval was issued to the given director node according to a clock of the worker node;
receiving, from the given director node, a response to the request for approval that identifies a second time at which the request for approval was received at the given director node according to the clock of the given director node; and
updating the delay information to include a time delay based on the first and second times.
13. The medium of claim 12 , wherein the operations further comprise:
including the delay information in the request for approval to enable a first director node of the director nodes to attempt to align a time observed by the first director node with a time observed by a second director node of the director nodes.
14. The medium of claim 10 , wherein the operations further comprise:
aborting the transaction in response to receiving, from at least one of the director nodes, a disapproval response that indicates a disapproval of the proposed commit time based on an arrival of the request at the at least one director node after the proposed commit time.
15. A method, comprising:
accessing, by a worker node of a database system, delay information specifying time delays in communicating with ones of a plurality of director nodes of the database system that are operable to ensure transactional consistency for transactions of the database system;
receiving, by the worker node, a request to perform a transaction;
selecting, by the worker node, a proposed commit time for the transaction based on the delay information and a clock of the worker node; and
issuing, by the worker node to two or more of the plurality of director nodes, a first request for approval to commit the transaction, wherein the first request specifies the proposed commit time, and wherein a given director node of the two or more director nodes is operable to process the first request for approval upon reaching the proposed commit time according to a clock of the given director node.
16. The method of claim 15 , further comprising:
prior to issuing the first request for approval, the worker node issuing a second request to the given director node for approval of a proposed time associated with another transaction;
recording, by the worker node, a first time at which the second request for approval was issued to the given director node according to the clock of the worker node; and
receiving, by the worker node from the given director node, a response to the second request for approval that identifies a second time at which the second request for approval was received at the given director node according to the clock of the given director node, wherein the clock of the worker node identifies a different time than the clock of the given director node, and wherein the proposed commit time is selected based on the first and second times.
17. The method of claim 15 , wherein the proposed commit time is a future time that is determined to permit the request for approval to arrive at the given director node prior to the future time according to the clock of the given director node.
18. The method of claim 15 , further comprising:
sending, by the worker node, the delay information to the director nodes to enable ones of the director nodes to attempt to align times identified by clocks of the director nodes.