Various updates

- Update docs
- Update azure-batch dependency
- Set Slurm scheduling option defer mode

CHANGELOG.md

@@ -23,8 +23,6 @@ been migrated under the new `additional_node_prep` property as
 - Performance improvements to speed up job submission with large task
 factories or large amount of tasks. Verbosity of task generation progress
 has been increased which can be modified with `-v`.
-- Updated Docker CE to 18.09.2
-- Updated Singularity to 2.6.1
 - Updated blobxfer to 1.7.0 ([#255](https://github.com/Azure/batch-shipyard/issues/255))
 - Updated LIS, NV driver to 410.92 and NC/ND driver to 410.104
 - Updated other dependencies to latest
@@ -35,9 +33,14 @@ supplied parameters ([#249](https://github.com/Azure/batch-shipyard/issues/249))
 - Azure Function extension installation failure ([#260](https://github.com/Azure/batch-shipyard/issues/260))
 - Block job submission on non-active pools ([#251](https://github.com/Azure/batch-shipyard/issues/251))
 - Missing files included in binary distributions ([#258](https://github.com/Azure/batch-shipyard/issues/258))
-- Pools with accelerated networking with fail provisioning sometimes due to
+- Pools with accelerated networking would fail provisioning sometimes due to
 infiniband devices being present for non-RDMA VM sizes
 
+### Security
+- Updated Docker CE to 18.09.2 to address the runc CVE-2019-5736
+- Updated Singularity to 2.6.1 to address the shared mount propagation
+vulnerability CVE-2018-19295
+
 ## [3.6.1] - 2018-12-03
 ### Added
 - `force_enable_task_dependencies` property in jobs configuration to turn

cargo/requirements.txt

@@ -1,3 +1,3 @@
-azure-batch==6.0.0
+azure-batch==6.0.1
 msrest==0.5.5
 requests==2.21.0

docs/64-batch-shipyard-byovnet.md

@@ -110,8 +110,9 @@ subnet level, then you will need to create a single `Inbound Security Rule`
 to allow traffic in at the Virtual Network level for Batch compute nodes to
 successfully operate.
 
-Ports `29876` and `29877` must allow `TCP` traffic from the Source
-`Service Tag` of `BatchNodeManagement` to `Any` Destination as shown below:
+Ports `29876` and `29877` must allow `TCP` traffic from the
+`Source service tag` of `BatchNodeManagement` to `Any` Destination as shown
+below:
 
 ![64-byovnet-nsg-inbound-rule.png](https://azurebatchshipyard.blob.core.windows.net/github/64-byovnet-nsg-inbound-rule.png)
 
@@ -137,7 +138,7 @@ restrict outbound network traffic, ensure that you have either the generic
 `Storage` service tag or the correct `Storage.<region>` service tag.
 Additionally, ensure that if you are specifying the destination port that
 you provide sufficient rules to cover all outbound requests over port `443`
-including potentially accesses to other storage regions or any application
+including potential requests to other storage regions or any application
 logic that may use port `443`.
 
 ## Additional Configuration Documentation

docs/69-batch-shipyard-slurm.md

@@ -319,5 +319,5 @@ separately. User subscription based Batch accounts share the underlying
 subscription regional core quotas.
 
 ## Sample Usage
-Please see the sample [Slurm recipe](../recipes/Slurm-NFS) for a working
+Please see the sample [Slurm recipe](../recipes/Slurm+NFS) for a working
 example.

federation/requirements.txt

@@ -1,4 +1,4 @@
-azure-batch==6.0.0
+azure-batch==6.0.1
 azure-cosmosdb-table==1.0.5
 azure-mgmt-compute==4.4.0
 azure-mgmt-resource==2.1.0

heimdall/requirements.txt

@@ -1,4 +1,4 @@
-azure-batch==6.0.0
+azure-batch==6.0.1
 azure-cosmosdb-table==1.0.5
 azure-mgmt-compute==4.4.0
 azure-mgmt-network==2.5.1

recipes/PyTorch-GPU/README.md

@@ -1,4 +1,4 @@
-# PyTorch-CPU
+# PyTorch-GPU
 This recipe shows how to run [PyTorch](https://pytorch.org/) on GPUs
 using N-series Azure VM instances in an Azure Batch compute pool.
 This sample executes the MNIST example.

recipes/README.md

@@ -130,7 +130,7 @@ This PyTorch-CPU recipe contains information on how to containerize
 [PyTorch](https://pytorch.org) for use on Azure Batch compute nodes.
 
 #### [PyTorch-GPU](./PyTorch-GPU)
-This Torch-GPU recipe contains information on how to containerize
+This PyTorch-GPU recipe contains information on how to containerize
 [PyTorch](https://pytorch.org) on GPUs for use with N-series Azure VMs.
 
 #### [TensorFlow-CPU](./TensorFlow-CPU)

requirements.txt

@@ -1,5 +1,5 @@
 adal==1.2.1
-azure-batch==6.0.0
+azure-batch==6.0.1
 azure-cosmosdb-table==1.0.5
 azure-keyvault==1.1.0
 azure-mgmt-authorization==0.51.1

slurm/requirements.txt

@@ -1,4 +1,4 @@
-azure-batch==6.0.0
+azure-batch==6.0.1
 azure-cosmosdb-table==1.0.5
 azure-mgmt-storage==3.1.1
 azure-mgmt-resource==2.1.0

slurm/slurm.conf

@@ -27,7 +27,6 @@ PluginDir=/usr/lib/slurm
 # LOGGING
 SlurmctldDebug=5
 # slurmctld log file has no %h option, so need to log locally
-#SlurmctldLogFile=/var/log/slurm-llnl/slurmctld.log
 SlurmctldLogFile=/var/log/slurm/slurmctld.log
 SlurmdDebug=5
 SlurmdLogFile={SLURM_LOG_PATH}/slurmd-%h.log
@@ -67,6 +66,7 @@ Waittime=0
 
 # SCHEDULING
 SchedulerType=sched/backfill
+SchedulerParameters=defer
 #SchedulerAuth=
 SelectType=select/linear
 #SelectType=select/cons_res