[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Vinod Kumar Vavilapalli updated YARN-569: - Issue Type: Bug (was: Sub-task) Parent: (was: YARN-397) CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Bug Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Fix For: 2.1.0-beta Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, YARN-569.1.patch, YARN-569.10.patch, YARN-569.11.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.5.patch, YARN-569.6.patch, YARN-569.8.patch, YARN-569.9.patch, YARN-569.patch, YARN-569.patch, preemption.2.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Vinod Kumar Vavilapalli updated YARN-569: - Issue Type: Sub-task (was: Bug) Parent: YARN-45 CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Fix For: 2.1.0-beta Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, YARN-569.1.patch, YARN-569.10.patch, YARN-569.11.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.5.patch, YARN-569.6.patch, YARN-569.8.patch, YARN-569.9.patch, YARN-569.patch, YARN-569.patch, preemption.2.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Douglas updated YARN-569: --- Fix Version/s: 2.1.0-beta CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Fix For: 2.1.0-beta Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.10.patch, YARN-569.11.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.5.patch, YARN-569.6.patch, YARN-569.8.patch, YARN-569.9.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Douglas updated YARN-569: --- Attachment: YARN-569.11.patch Rebase. CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.10.patch, YARN-569.11.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.5.patch, YARN-569.6.patch, YARN-569.8.patch, YARN-569.9.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Douglas updated YARN-569: --- Attachment: YARN-569.10.patch CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.10.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.5.patch, YARN-569.6.patch, YARN-569.8.patch, YARN-569.9.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Douglas updated YARN-569: --- Attachment: YARN-569.9.patch bq. One other thing to check would be if the preemption policy will use refreshed values when the capacity scheduler config is refreshed on the fly. Looks like cloneQueues() will take the absolute used and guaranteed numbers on every clone. So we should be good wrt that. Would be good to check other values the policy looks at. *nod* Right now, the policy rebuilds its view of the scheduler at every pass, but it doesn't refresh its own config parameters. bq. Noticed formatting issues with spaces in the patch. eg. cloneQueues() Did another pass over the patch, fixed up spacing, formatting, and removed obvious whitespace changes. Sorry, did a few of these already, but missed a few. Also moved the check in the {{ApplicationMasterService}} as part of this patch. CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.5.patch, YARN-569.6.patch, YARN-569.8.patch, YARN-569.9.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Douglas updated YARN-569: --- Attachment: YARN-569.8.patch CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.5.patch, YARN-569.6.patch, YARN-569.8.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Douglas updated YARN-569: --- Attachment: YARN-569.6.patch CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.5.patch, YARN-569.6.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Douglas updated YARN-569: --- Attachment: YARN-569.4.patch Rebase after YARN-635, YARN-735, YARN-748, YARN-749. Fixed findbugs warnings. CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.4.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Chris Douglas updated YARN-569: --- Attachment: YARN-569.3.patch CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.3.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect balance by some small % we ignore it) # overall amount of
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Carlo Curino updated YARN-569: -- Attachment: YARN-569.2.patch CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.1.patch, YARN-569.2.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect balance by some small % we ignore it) # overall amount of preemption we can
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Carlo Curino updated YARN-569: -- Attachment: YARN-569.1.patch CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.1.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect balance by some small % we ignore it) # overall amount of preemption we can afford for each run of
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Bikas Saha updated YARN-569: Attachment: preemption.2.patch Attaching a patch that contains wip code to add preemption to the capacity scheduler. It was written pre-DRF times. The approach is similar to the current efforts in having the logic in a separate thread. So most of the code should still easily apply. The approach differs in that it turns off reservation and also specifies where the preempted resources should go. Hopefully there will be something helpful in it to contribute to the efforts in this jira. CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, preemption.2.patch, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Carlo Curino updated YARN-569: -- Attachment: YARN-569.patch CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, YARN-569.patch, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect balance by some small % we ignore it) # overall amount of preemption we can afford for each run of the policy (in terms of total cluster
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Carlo Curino updated YARN-569: -- Attachment: YARN-569.patch CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect balance by some small % we ignore it) # overall amount of preemption we can afford for each run of the policy (in terms of total cluster capacity) In our
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Carlo Curino updated YARN-569: -- Attachment: (was: capacity.patch) CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Sub-task Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, CapScheduler_with_preemption.pdf, YARN-569.patch There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect balance by some small % we ignore it) # overall amount of preemption we can afford for each run of the policy (in terms of total cluster
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Carlo Curino updated YARN-569: -- Attachment: 3queues.pdf CapScheduler_with_preemption.pdf CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Improvement Components: capacityscheduler Reporter: Carlo Curino Attachments: 3queues.pdf, capacity.patch, CapScheduler_with_preemption.pdf There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect balance by some small % we ignore it) # overall amount of preemption we can afford for each run of the policy (in terms of total cluster
[jira] [Updated] (YARN-569) CapacityScheduler: support for preemption (using a capacity monitor)
[ https://issues.apache.org/jira/browse/YARN-569?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Carlo Curino updated YARN-569: -- Assignee: Carlo Curino CapacityScheduler: support for preemption (using a capacity monitor) Key: YARN-569 URL: https://issues.apache.org/jira/browse/YARN-569 Project: Hadoop YARN Issue Type: Improvement Components: capacityscheduler Reporter: Carlo Curino Assignee: Carlo Curino Attachments: 3queues.pdf, capacity.patch, CapScheduler_with_preemption.pdf There is a tension between the fast-pace reactive role of the CapacityScheduler, which needs to respond quickly to applications resource requests, and node updates, and the more introspective, time-based considerations needed to observe and correct for capacity balance. To this purpose we opted instead of hacking the delicate mechanisms of the CapacityScheduler directly to add support for preemption by means of a Capacity Monitor, which can be run optionally as a separate service (much like the NMLivelinessMonitor). The capacity monitor (similarly to equivalent functionalities in the fairness scheduler) operates running on intervals (e.g., every 3 seconds), observe the state of the assignment of resources to queues from the capacity scheduler, performs off-line computation to determine if preemption is needed, and how best to edit the current schedule to improve capacity, and generates events that produce four possible actions: # Container de-reservations # Resource-based preemptions # Container-based preemptions # Container killing The actions listed above are progressively more costly, and it is up to the policy to use them as desired to achieve the rebalancing goals. Note that due to the lag in the effect of these actions the policy should operate at the macroscopic level (e.g., preempt tens of containers from a queue) and not trying to tightly and consistently micromanage container allocations. - Preemption policy (ProportionalCapacityPreemptionPolicy): - Preemption policies are by design pluggable, in the following we present an initial policy (ProportionalCapacityPreemptionPolicy) we have been experimenting with. The ProportionalCapacityPreemptionPolicy behaves as follows: # it gathers from the scheduler the state of the queues, in particular, their current capacity, guaranteed capacity and pending requests (*) # if there are pending requests from queues that are under capacity it computes a new ideal balanced state (**) # it computes the set of preemptions needed to repair the current schedule and achieve capacity balance (accounting for natural completion rates, and respecting bounds on the amount of preemption we allow for each round) # it selects which applications to preempt from each over-capacity queue (the last one in the FIFO order) # it remove reservations from the most recently assigned app until the amount of resource to reclaim is obtained, or until no more reservations exits # (if not enough) it issues preemptions for containers from the same applications (reverse chronological order, last assigned container first) again until necessary or until no containers except the AM container are left, # (if not enough) it moves onto unreserve and preempt from the next application. # containers that have been asked to preempt are tracked across executions. If a containers is among the one to be preempted for more than a certain time, the container is moved in a the list of containers to be forcibly killed. Notes: (*) at the moment, in order to avoid double-counting of the requests, we only look at the ANY part of pending resource requests, which means we might not preempt on behalf of AMs that ask only for specific locations but not any. (**) The ideal balance state is one in which each queue has at least its guaranteed capacity, and the spare capacity is distributed among queues (that wants some) as a weighted fair share. Where the weighting is based on the guaranteed capacity of a queue, and the function runs to a fix point. Tunables of the ProportionalCapacityPreemptionPolicy: # observe-only mode (i.e., log the actions it would take, but behave as read-only) # how frequently to run the policy # how long to wait between preemption and kill of a container # which fraction of the containers I would like to obtain should I preempt (has to do with the natural rate at which containers are returned) # deadzone size, i.e., what % of over-capacity should I ignore (if we are off perfect balance by some small % we ignore it) # overall amount of preemption we can afford for each run of the policy (in terms of total cluster capacity) In our