PACE A Partnership for an Advanced Computing Environment

October 27, 2020

Hive Scratch Storage Update

Filed under: Uncategorized — Michael Weiner @ 10:39 am

We would like to remind you about scratch storage policy on Hive. Scratch is designed for temporary storage and is never backed up. Each week, files not modified for more than 60 days are automatically deleted from your scratch directory. As part of Hive’s start-up, regular cleanup of scratch has now been implemented. Each week, users with files set to be deleted receive a warning email about files to be deleted in the coming week, with additional information included. Those of you who use the main PACE system are already familiar with this workflow.

Some of you received such an email yesterday. As always, if you need additional time to migrate valuable data off of scratch, please respond to the email as directed to request a delay.

Please contact us at pace-support@oit.gatech.edu with any questions about how to manage your data stored on Hive.

October 26, 2020

CoE HPC Cost Model Listening Session

Filed under: Uncategorized — Semir Sarajlic @ 7:15 pm

Over the past few months, a team from the EVPR, OIT, EVP-A&F, and GTRC has been working with Institute leadership to develop a more sustainable and flexible way to support research cyberinfrastructure. This new model is described in more detail below and will affect researchers who leverage PACE services. The model enjoys strong support, but it is not yet fully approved.  We are communicating at this stage because we wanted you to be aware of the upcoming changes and we welcome your feedback. Please submit comments to the PACE Team <pace-support@oit.gatech.edu> or to Lew Lefton <lew.lefton@gatech.edu>. This listening session is organized for the College of Engineering.

Date:           11/02/2020, 4:00pm – 5:00pm

Location:   BlueJeans (link provided via email)

Host:           EVPR/PACE

In a nutshell, PACE will transition from a service that purchases nodes with equipment funds, to a service which operates as a Cost Center. This means that major research cyberinfrastructure (including compute and storage services) will be treated like other core facilities. This new model will begin as the transition to the new equipment in the CODA data center happens. We recognize that this represents a shift in how we think about research computing. But, as shown below, the data indicates that the long-term benefits are worth the change.  When researchers only pay for actual consumption – similar to commercial cloud offerings from AWS, Azure, and GCP – there are several advantages:

  • Researchers have more flexibility to leverage new hardware releases instead of being restricted to hardware purchased at a specific point in time.
  • The PACE team can use capacity and usage planning to make compute cycles available to faculty in days or week as opposed to having to wait for months due to procurement bottlenecks.
  • We have secured an Indirect Cost Waiver on both PACE services and commercial cloud offerings for two years to allow us to collect data on the model and see how it is working.
  • Note that a similar consumption model has been used successfully at other institutions such as Univ. Washington and UCSD, and this approach is also being developed by key sponsors (e.g. NSF’s cloudbank.org).
  • A free tier that provides any PI the equivalent of 10,000 CPU-hours on a 192GB compute node and 1 TB of project storage at no cost.

For further details on the new cost model, please visit out Web page

October 13, 2020

CoS HPC cost model listening session

Filed under: Uncategorized — Semir Sarajlic @ 1:08 am

Over the past few months, a team from the EVPR, OIT, EVP-A&F, and GTRC has been working with Institute leadership to develop a more sustainable and flexible way to support research cyberinfrastructure. This new model is described in more detail below and will affect researchers who leverage PACE services. The model enjoys strong support, but it is not yet fully approved.  We are communicating at this stage because we wanted you to be aware of the upcoming changes and we welcome your feedback. Please submit comments to the PACE Team <pace-support@oit.gatech.edu> or to Lew Lefton <lew.lefton@gatech.edu>. This listening session is organized for the College of Sciences.

Date:           10/13/2020, 10:00am – 11:00am

Location:   BlueJeans (link provided via email)

Host:           EVPR/PACE

In a nutshell, PACE will transition from a service that purchases nodes with equipment funds, to a service which operates as a Cost Center. This means that major research cyberinfrastructure (including compute and storage services) will be treated like other core facilities. This new model will begin as the transition to the new equipment in the CODA data center happens. We recognize that this represents a shift in how we think about research computing. But, as shown below, the data indicates that the long-term benefits are worth the change.  When researchers only pay for actual consumption – similar to commercial cloud offerings from AWS, Azure, and GCP – there are several advantages:

  • Researchers have more flexibility to leverage new hardware releases instead of being restricted to hardware purchased at a specific point in time.
  • The PACE team can use capacity and usage planning to make compute cycles available to faculty in days or week as opposed to having to wait for months due to procurement bottlenecks.
  • We have secured an Indirect Cost Waiver on both PACE services and commercial cloud offerings for two years to allow us to collect data on the model and see how it is working.
  • Note that a similar consumption model has been used successfully at other institutions such as Univ. Washington and UCSD, and this approach is also being developed by key sponsors (e.g. NSF’s cloudbank.org).
  • A free tier that provides any PI the equivalent of 10,000 CPU-hours on a 192GB compute node and 1 TB of project storage at no cost.

For further details on the new cost model, please visit out Web page

October 6, 2020

[Resolved] Power Outage at Rich Datacenter

Filed under: Uncategorized — Semir Sarajlic @ 4:54 pm
[Update – 10/07/2020 – 8:02]
After nearly-28 hours since the initial power outage in the Rich Datacenter that further caused complications and failures with the networks and systems, we are pleased to report that we have restored the PACE resources in Rich Datacenter and released the user jobs.   We understand the impact this has had on your research, and we are very grateful for your patience and understanding as we worked through this emergency.  During this outage, the PACE clusters in the Coda datacenter (Hive, Testflight-Coda, CoC-ICE, PACE-ICE, and Phoenix) have not been impacted.
What we have done:  Since last night after the network repairs were conducted, we were closely monitoring the network/fabric, and we have gradually brought the infrastructure back up.  We conducted application and fabric testing across the systems to assure the systems are operational, and we addressed problematic nodes and issues with schedulers.  The power and fabric are stable. We have identified the users whose jobs were interrupted by this power outage from yesterday, and we will reach out to impacted users directly.  We have released user jobs that were queued prior to the power outage when we paused the schedulers, and jobs are currently running.
What we will continue to do: PACE team will continue to monitor the systems, and we will report as needed.      We have some straggling nodes that will remain offline, and we will work to bring them back up in the coming days.
Please don’t hesitate to contact us at pace-support@oit.gatech.edu if you have any questions or if you encounter any issues on the clusters.  Thank you again for your patience.
[Update – 10/06/2020 – 11:20]

We are following up to update you on the current status of the Rich Datacenter.    After a tireless evening, the PACE team in collaboration with OIT have successfully restored the network at approximately 11:00pm.  We replaced a failed management module on the core InfiniBand switch, now, the switch is operational.  Preliminary spot checks indicate that the fabric is stable.   In abundance of caution, we will monitor the network overnight.  In the morning, we aim to conduct additional testing and online the compute resources in Rich Datacenter, followed by releasing user jobs that are currently paused.    The power remains stable after the repairs were conducted, and the UPS is back at nearly full charge.

As always, thank you for your patience and understanding during this outage as we know how critical these resources are to your research.   

If you have any questions or concerns, please do not hesitate to contact us at pace-support@oit.gatech.edu.

[Update – 10/06/2020 – 6:30]

This is brief update on the current power outage.   Power has been restored in Rich datacenter, and recovery is underway.  Some Ethernet network switches have failed, and replacements and re-configurations are underway to try and restore services.  Currently, our core InfiniBand switch has not restarted yet.  We will continue to update you as we have more information.  For up to date information, please check the status and blog pages:

Again, this emergency work does not impact any of the resources in CODA datacenter.

Thank you for your continued patience and understanding as we work through this emergency.

[Original Post – 10/06/2020 – 4:54] 
We have a power outage on a section of campus that includes the Rich datacenter’s 133 computer room.  We are urgently shutting down the schedulers and remaining servers in Rich133.  Power to storage and login nodes in Rich are currently on generator power and will remain safe.
What is happening and what we have done:  At 3:45pm the campus (not GA Power) distribution power issued a power outage, and at 4:05 Rich 133 UPS went out.  Power to the chillers and to 2/3 of the computer room in Rich Datacenter is out. .  Facilities is on site and investigating the situation, also, High Voltage contractor is in route. We have initiated urgent shutdown of schedulers and remaining servers in the Rich datacenter’s 133 computer room.   Storage and login nodes are running on generators, but most of the running user jobs will have been interrupted by this power outage.

What we will continue to do: This is an active situation, and we will follow up with updates as they become available, and for most up to date information, please check the status and blog pages:

This emergency work does not impact any of the resources in CODA datacenter.

Thank you for your attention to this urgent message, and we apologize for this inconvenience.

The PACE Team

October 5, 2020

[RESOLVED] URGENT – CODA datacenter research hall emergency shutdown

Filed under: Uncategorized — Semir Sarajlic @ 6:16 pm

[Update – 10/05/2020 8:18]

Thank you for your patience as we worked through this emergency to restore cooling in the CODA datacenter’s Research Hall.  At this time, we have Hive, COC-ICE, PACE-ICE, Testflight-CODA and Phoenix clusters back online with users’ previously queued jobs having started.

What has happened and what we did:   At 4:30pm today, the main chiller for the research computing failed fully in CODA datacenter’s Research Hall side.  PACE had urgently shutdown the compute nodes for the Hive, COC-ICE, PACE-ICE, Testflight-CODA and Phoenix clusters.  Storage and login nodes were not impacted during this outage.  Working with DataBank, we were able to restore enough cooling  using economizer module that can handle all cooling in the Research Hall.   At 6:30pm, we had onlined Hive cluster, and since then we have continued to bring back up the remaining cluster’s compute nodes for COC-ICE, PACE-ICE, Testflight-CODA, and Phoenix clusters while maintaining normal operating temperatures.   At about 7:00pm vendor has arrived, and is working on chiller, and no interruption should occur when the fixed chiller is brought online.  Our storage did not experience data loss, but users’ running jobs were interrupted by this emergency shutdown.  We encourage users to check on their jobs and resubmit any jobs that may have been interrupted.  Currently, previously queued user jobs are running on the clusters.

What we will continue to do:   PACE team continue to monitor the situation, and report accordingly as needed.

For your reference we are including OIT’s status page link and blog post:

Status page:  https://status.gatech.edu/pages/incident/5be9af0e5638b904c2030699/5f7b9062cb294e04bbe8cbda

Blog post: https://blog.pace.gatech.edu/?p=6931

If you have any questions or concerns, please direct them to pace-support@oit.gatech.edu.

Thank you for your patience and attention to this emergency.

 

[Original Post – 10/05/2020 6:16]

The cooling has failed in CODA datacenter’s research hall.  We have initiated and completed emergency shutdown of all resources in CODA research hall that includes: Hive, COC-ICE, PACE-ICE, Testflight-CODA, and the Phoenix clusters.

What is happening and what we have done:   We have urgently  completed emergency shutdown of all the clusters in CODA datacenter.  Research data and cluster headnodes are fine, but all running user jobs will have been interrupted by this outage.  At this time, we are using economizer module to provide some cooling, and we are beginning to bring back up Hive cluster while closely monitoring the temperatures.

What we will continue to do: This is an active situation, and we will follow up with updates as they become available.

Also, please follow the updates on the OIT’s status page: https://status.gatech.edu/pages/incident/5be9af0e5638b904c2030699/5f7b9062cb294e04bbe8cbda

Additionally, we are tracking the updates in our blog at: https://blog.pace.gatech.edu/?p=6931

This emergency work does not impact any of the resources in Rich datacenter.

Thank you for your attention to this urgent message.

 

 

October 3, 2020

[Resolved] TestFlight-Coda, COC-ICE, and PACE-ICE Schedulers Down

Filed under: Uncategorized — Semir Sarajlic @ 1:43 pm

[Update – 10/05/2020 – 10:20am]

PACE has completed testing across the resources in Coda datacenter over the weekend.  These tests did not impact Hive cluster or PACE resources in Rich datacenter.  We brought the schedulers for coc-ice, pace-ice, and Testflight-coda clusters online.  Users queued jobs have resumed.

If you have any questions, please don’t hesitate to contact us at pace-support@oit.gatech.edu.

Thank you again for your patience during this testing.

[Original Post – 10/03/2020 – 1:43pm]

In order to complete preparations for bringing PACE’s new Coda resources on the research cluster into production, we had to urgently offline testflight-coda, coc-ice, and pace-ice schedulers on Saturday at about 10:30am, which is in effect until 8 AM on Monday. We did have a job reservation in place to prevent interruptions to any user jobs, and at the time of taking testflight-coda, coc-ice, and pace-ice schedulers offline, there were no users running jobs on the system.  We apologize for this inconvenience. You can still access the login node over the weekend, but you will receive an error message if you attempt to submit a job.  Your files are all accessible via the login node.  All queued jobs prior to the offlining of the schedulers will resume on Monday.

Hive and all PACE resources in the Rich datacenter are not affected.

Again, we apologize for the late notice. Please contact us at pace-support@oit.gatech.edu with questions.

Powered by WordPress