[0111] Wherein the pressure on the hair for hair pressure node satisfies the termination condition is determined accurately, a plurality of hair-dividing node satisfies the termination condition of the hair pressing an alternative embodiment as follows:
[0112] The number of nodes history of at least one target node is used when a stress test before the acquisition.
[0113] When the total number of nodes is less than the history of a plurality of nodes send pressure nodes, determining a plurality of hair pressing pressure nodes send termination condition is not satisfied.
[0114] When the number of nodes equal to the total amount of history node of a plurality of hair-dividing node, determining a plurality of hair pressing pressure node satisfies the termination condition hair.
[0115] Total pressure node may send a plurality of nodes the number of participating nodes may send all the hair-dividing node pressure, when a plurality of nodes can send hair pressure press, the number of nodes that can send a plurality of pressure nodes.
[0116] In this embodiment, the number of nodes on the hair using a termination condition is a plurality of hair pressing pressure for accurate determination nodes, number of nodes is determined using a direct manner, simple, accurate, high efficiency, and determine the accuracy determination.
[0117] Among them, the resource usage information can be sent to multiple pressure nodes in a cluster stress test is currently involved in the use of hardware resources of the data target node test.
[0118] Specifically, the resource use information may include: at least once a stress test prior usage history per a certain destination node, the central processor utilization, the memory of at least one.
[0119] Specific contents of this embodiment, the throughput of the at least a second target node, the CPU usage rate and the memory usage and other information using the precise definition of the information resources, the resource use information specific to the present embodiment, in order to utilize more detail resource usage information for subsequent node is determined, the exact conditions to achieve the step of determining the like, to further improve the precision of the protection provided.
[0120] The number of resource usage information and history nodes can simultaneously send multiple pressure nodes used to measure whether the current pressure termination condition hair. When the total number of nodes is less than the history of a plurality of nodes send pressure nodes, determining a plurality of hair pressing pressure nodes send termination condition is not satisfied. When the number of nodes equal to the total amount of history node of a plurality of hair-dividing node, may utilize the resource use information further determines whether a plurality of hair pressing pressure node satisfies the termination condition hair.
[0121] Wherein, when the number of nodes equal to the history of a plurality of nodes issued total pressure node, determining a plurality of hair pressing pressure node satisfies the termination condition may hair comprising: when the history of the number of nodes equal to the total plurality of nodes send pressure node, a plurality of throughput per node hair pressure less than the pressure measurement is integrated to produce the required service QPS (query per second, queries per second) and memory or CPU utilization rates for all hair pressure node is greater than 85% of the total resources, determining a plurality of hair pressing pressure node satisfies the termination condition hair.
[0122] When the number of nodes equal to the total amount of history node of the plurality of nodes send pressure, throughput per second plurality of nodes send pressure greater than the pressure generated in the integrated measurement of service requirements and utilization of all hair QPS pressure node CPU or memory is not full greater than 85% of total resources, determining a plurality of hair pressing pressure nodes send termination condition is not satisfied, the task can continue to increase the amount of pressure measurement, pressure measurement task to continue. Especially for concurrent test scenarios, you may continue to perform in the case of concurrent treatment measurement test server, and transmits the number of concurrent testing step is determined.
[0123]History of at least one target node node used in a stress test before getting to the number, the current number of nodes can be calculated according to the number of the history node. Thus, Step 504: In a case where a plurality of nodes issued press pressure hair termination condition is not satisfied, the use of resource use information, determining the number of nodes participating node hair pressure stress tests may include:
[0124] In the case where a plurality of nodes issued press pressure does not satisfy the termination condition of the hair, using the resource use information to determine the need to increase the node increments.
[0125] The history of computing nodes and node number of increments and get the number of nodes.
[0126] Alternatively, when the number of nodes is less than the total amount of history node of the plurality of nodes send pressure, pressure may be determined plurality of nodes send the termination condition is not satisfied send pressure. Resource usage information may include: at least one usage history per second throughput target node, the CPU usage, memory during a stress test before at least one. When the total number of nodes is less than the history of a plurality of nodes send pressure nodes using the resource use information to determine the need to increase the number of nodes may include: when the number of nodes is less than the total amount of history node send a plurality of pressure nodes, if at least one historical when the target node reaches per second throughput which is smaller than would result from QPS (query per second, queries per second) and pressure measurement service requirement memory or CPU utilization rates of all the nodes send pressure greater than 85% of all resources, calculating at least one target node's history and the history of the number of nodes and node increments of, get the number of nodes.
[0127] The node increments may set the actual needs, each test the number of nodes increased. For example, node 1 may be set to increments.
[0128] In this embodiment, when a plurality of hair pressing pressure nodes send the termination condition is not satisfied, the use of resource use information, determining the number of nodes of the at least one target node send a plurality of pressure nodes, such that the number of the resource node information related to use of hardware resources, the number of nodes to obtain more accurate, pressure testing at least one target node corresponding to the node number, the actual flow test request generated from impact of hardware resources, hardware resources is not enough to avoid pressure error results in hair, improve the precision.
[0129] In the actual test process may be employed concurrently testing approach to test server cycle test was performed, from the smaller tasks amount starts increasing assignments, service flow is not para to send pressure node sends increase, accurate treatment test service side test.
[0130] like Image 6 , The present disclosure fourth flow chart of a stress test method according to an embodiment, the method may comprise the following steps:
[0131] 601: determining the target server treat stress test test test data.
[0132] 602: obtaining a plurality of cluster resources stress test voltage nodes send corresponding hardware resources usage information.
[0133] 603: In the case of concurrent treatment test server test, to determine the number of concurrent concurrent testing.
[0134] 604: In the case where a plurality of nodes issued press pressure does not satisfy the termination condition of the hair, the use of resource use information, determining the number of nodes participating node hair pressure stress tests.
[0135] 605: The number of nodes, and the number of concurrent tasks to determine the amount of concurrent tasks are performed at least one target node.
[0136] Alternatively, assignments correspond to the number of threads may be started target node, each thread starts, the test data generated according to the target service request.
[0137] Tasks amount of concurrent tasks to perform at least one target node, respectively, can be determined based on the number of concurrent and number of nodes quotient and remainder. In particular, the number of nodes is k, the number of concurrent is n, for example, for the k-th target node in the i-th target node, if i> n% k (the number of concurrent with the number of nodes in the remainder), it is necessary the i-th target the amount of task execution node is n / k (the number of nodes and the number of concurrent commercially rounded), if i
[0138] 606: The number of nodes and the amount of the at least one task target node performs concurrent tasks, respectively, to generate the target node information.
[0139] Wherein the information for the destination node determines the control node for the at least one target node from the plurality of nodes according to the number of hair pressure nodes; test data using the target control at least one destination node in accordance with their respective assignments, to initiate a test server stress test.
[0140] Alternatively, the number of tasks in accordance with the amount of nodes and at least one target node performs concurrent tasks, respectively, to generate at least one node information of the target node may include: obtaining the amount of concurrent tasks in the task at least one target node are performed to obtain an amount of at least one task , the number of nodes and at least one package an amount of task information to generate the target node. After the target node receives the control node information, number of nodes, and can obtain an amount of the at least one task in the destination node information, the number of nodes, selecting at least one target node from the plurality of nodes send pressure, and the amount of the at least one task, sequentially distributed to the at least one target node, the target node for each amount to obtain the corresponding task, and initiates stress tests according to the corresponding test server assignments.
[0141] Alternatively, a target node to initiate any stress testing in accordance with their respective assignments to the boot server may specifically be a corresponding number of threads or processes in accordance with the amount of tasks to obtain a plurality of threads or processes, each generating at least one test data according to a target service request , all service requests to all threads or processes generated so as to simultaneously transmit the request to the stream server under test.
[0142] 607: a destination node information and test data to the control target node cluster stress test to indicate the control node using the target control at least one test data according to their respective destination node assignments, to initiate a test server stress test.
[0143] Alternatively, the amount of concurrent tasks task at least one target node may be performed separately determined by the control node. After the control node determines the at least one target node, depending on the number of nodes, and the number of concurrent tasks to determine the amount of concurrent tasks are performed at least one target node. When determining at least one target node, the task may be determined according to the amount of the above-described step 605, the method described in detail, to determine assignments of each destination node, assign tasks to achieve accurate amount.
[0144] In this embodiment, the embodiment may be used to treat concurrent test test server stress test. By increasing the number of concurrent test Concurrency can achieve accurate test measurement services to treat end, the control node multiple concurrent. Use of concurrent test mode, you can achieve rapid concurrent. Improve test efficiency and test accuracy.
[0145] To treat an accurate test measurement server, the server can be treated measurement straight concurrent test cycle. Each time the number of concurrent test test may be determined by the number of concurrent test history. As a possible implementation, in the case of concurrent treatment test server test, to determine the number of concurrent concurrent testing, comprising:
[0146] In the case of concurrent treatment test server test, the number of concurrent history before getting a concurrent testing;
[0147] Depending on the number of concurrent history, as well as concurrent preset increments, determine the number of concurrent concurrent testing.
[0148] The number of concurrent history can be complicated by the number of concurrent and incremental.
[0149] In this embodiment, to set the concurrent increments, the number of concurrent concurrent test for accurate calculation, the number of concurrent improve extraction efficiency and accuracy.
[0150] Alternatively, the number of concurrent also be obtained by multiplying the number of concurrent and according to the number of concurrent single.
[0151] In practice, you can set the maximum number of concurrent. The maximum number of concurrent by a concurrent condition determination may be performed prior to concurrent test is greater than the maximum number of concurrent, after greater than the maximum number of concurrent, concurrent stop test. Greater than the maximum number of concurrent number of concurrent may be equal to or less than the number of concurrent maximum number of concurrent.
[0152] As an embodiment, the method further comprising:
[0153] Get maximum number of concurrent and concurrent increment set by the user;
[0154] The number of nodes, and the number of concurrent, determining at least the amount of concurrent tasks are tasks executed a target node, comprising:
[0155] In the case where the number of concurrent does not exceed the maximum number of concurrent, according to the number of nodes, and the number of concurrent tasks to determine the amount of concurrent tasks are performed at least one target node.
[0156] In this embodiment, by monitoring the number of concurrent accurate, an accurate distribution can be achieved concurrent task, the task of improving the accuracy of the distribution, to avoid inefficient distribution.
[0157] As an alternative, in order to accurately obtain the number of nodes and the number of concurrent. like Figure 7 As shown in the above embodiments, the resource use information obtaining step 502 a pressure test a plurality of cluster nodes send pressure corresponding to the hardware resources may include:
[0158] 701: Get at least a historical resource target node when a stress test before the pressure test cluster usage information.
[0159] This information may include resource usage: per second throughput, the CPU usage, memory usage rate of at least one.
[0160] After performing step 701, execution may be obtained in the foregoing embodiments, the pressure is determined whether a plurality of nodes send termination condition is satisfied, the following specific embodiments:
[0161] 702: determining the number of nodes at least one node of the historical development has reached the total pressure of the node, if so, step 703 is performed; if not, step 708 is performed.
[0162] 703: determining at least one historical whether the target node per second throughput which would result from the measured pressure is less than a QPS of service requirements and the use rate of a CPU or memory of all the nodes send pressure greater than 85% of all resources, if yes, step 704 ~ 705; if not, steps 706 ~ 707.
[0163] Determining the participating nodes send pressure stress tests of an embodiment of the node number as follows:
[0164] 704: calculating at least one target node's history and the history of the number of nodes and node increments of, get the number of nodes.
[0165] Determining the number of concurrent concurrent test of an embodiment as follows:
[0166] 705: participation in determining the number of concurrent stress tests can be complicated by the number of the previous history of concurrent tests.
[0167] In the case of at least one hardware resource use history to reach the target node limit, it can increase the number of nodes, and re-execute the last time the pressure measurement tasks.
[0168] Nodes involved in determining pressure stress tests made of the number of nodes of a further embodiment as follows:
[0169] 706: determining at least one historical destination node number is the number of nodes nodes history.
[0170]Yet another embodiment of the concurrent quantity of concurrent testing is as follows:
[0171] 707: Determine the concurrent quantity of concurrent testing based on historical concurrency, and preset concurrency increments.
[0172] In the case where the hardware resources of at least one historical target node do not reach the limit, the pressure task can be increased, and the pressure compression function of the pressure regulation node can be further developed.
[0173] 708: Judging whether at least one historical target node reaches the throughput of the throughput of each of its integrated to less than QPS required for the comparative service requirements and all the CPUs of all pressure regulating nodes are greater than 85% of all resources, if yes, execution Step 709; if no, step 707 is performed.
[0174] 709: Stopping the tested server to perform pressure testing and outputting the prompt information that the pressure test failed.
[0175] After step 707 or step 705, the end, after the number of nodes and the quantity amount, the other steps can be disclosed, for example, executing the number of nodes, and concurrent numbers, determining at least one target node, respectively, performing the concurrent task, respectively. . It is also possible to perform the steps of judging whether the concurrent number is greater than the maximum concurrency.