Ensure ABT_THREAD_STACKSIZE>=18432 for dcpm storage class

dcpm storage class with NDCTL enabled requires at least 18432 stack size for ULT. PR-repos: pmdk@PR-38:11 Priority: 2 Cancel-prev-build: false Skip-nlt: false Skip-unit-tests: false Skip-func-test-vm: false Skip-test-rpms: false Allow-unstable-test: true Skip-func-hw-test-large: true Skip-func-hw-test-medium: false Skip-func-hw-test-medium-verbs-provider: true Required-githooks: true Signed-off-by: Tomasz Gromadzki <[email protected]>
daos-stack · Aug 28, 2024 · 150978c · 150978c
1 parent e3ed8f2
commit 150978c
Show file tree

Hide file tree

Showing 4 changed files with 22 additions and 0 deletions.
diff --git a/src/control/server/engine/config.go b/src/control/server/engine/config.go
@@ -9,6 +9,7 @@ package engine
 import (
 	"fmt"
 	"os"
+	"strconv"
 	"strings"
 
 	"github.com/pkg/errors"
@@ -28,6 +29,8 @@ const (
 	envLogMasks      = "D_LOG_MASK"
 	envLogDbgStreams = "DD_MASK"
 	envLogSubsystems = "DD_SUBSYS"
+
+	MIN_ABT_THREAD_STACKSIZE_FOR_DCPM = 18432
 )
 
 // FabricConfig encapsulates networking fabric configuration.
@@ -343,6 +346,18 @@ func (c *Config) Validate() error {
 		return errors.Wrap(err, "validate engine log subsystems")
 	}
 
+	// ensure 18k ABT stack size for engine with DCPM storage class
+	if c.Storage.Tiers[0].IsSCM() && c.Storage.Tiers[0].Class == storage.ClassDcpm {
+		stacksize, err := c.GetEnvVar("ABT_THREAD_STACKSIZE")
+		if err == nil {
+			i, err := strconv.Atoi(stacksize)
+			if err == nil && i >= MIN_ABT_THREAD_STACKSIZE_FOR_DCPM {
+				return nil
+			}
+		}
+		return errors.Errorf("env_var ABT_THREAD_STACKSIZE=%d is required for `dcpm` storage class", MIN_ABT_THREAD_STACKSIZE_FOR_DCPM)
+	}
+
 	return nil
 }
 

diff --git a/utils/config/examples/daos_server_tcp.yml b/utils/config/examples/daos_server_tcp.yml
@@ -38,6 +38,7 @@ engines:
     env_vars:                 # influence DAOS I/O Engine behavior by setting env variables
       - FI_SOCKETS_MAX_CONN_RETRY=1
       - FI_SOCKETS_CONN_TIMEOUT=2000
+      - ABT_THREAD_STACKSIZE=18432
 
     # Storage definitions (one per tier)
     storage:

diff --git a/utils/config/examples/daos_server_ucx.yml b/utils/config/examples/daos_server_ucx.yml
@@ -43,6 +43,9 @@ engines:
     log_mask: ERR
     log_file: /tmp/daos_engine.0.log
 
+    env_vars:
+      - ABT_THREAD_STACKSIZE=18432
+
     # Storage definitions (one per tier)
     storage:
       -

diff --git a/utils/config/examples/daos_server_verbs.yml b/utils/config/examples/daos_server_verbs.yml
@@ -34,6 +34,9 @@ engines:
     log_mask: ERR
     log_file: /tmp/daos_engine.0.log
 
+    env_vars:
+      - ABT_THREAD_STACKSIZE=18432
+
     # Storage definitions (one per tier)
     storage:
       -