diff --git a/cluster/log-dump/log-dump.sh b/cluster/log-dump/log-dump.sh
index 62c8eb2a9e1..f71d5fcd1a9 100755
--- a/cluster/log-dump/log-dump.sh
+++ b/cluster/log-dump/log-dump.sh
@@ -504,10 +504,10 @@ function dump_nodes_with_logexporter() {
   # Store logs from logexporter pods to allow debugging log exporting process
   # itself.
   proc=${max_dump_processes}
-  "${KUBECTL}" get pods -n "${logexporter_namespace}" -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{.spec.nodeName}{"\n"}{end}' | while read pod node; do
+  "${KUBECTL}" get pods -n "${logexporter_namespace}" -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{.spec.nodeName}{"\n"}{end}' | (while read -r pod node; do
     echo "Fetching logs from ${pod} running on ${node}"
-    mkdir -p ${report_dir}/${node}
-    "${KUBECTL}" logs -n "${logexporter_namespace}" ${pod} > ${report_dir}/${node}/${pod}.log &
+    mkdir -p "${report_dir}/${node}"
+    "${KUBECTL}" logs -n "${logexporter_namespace}" "${pod}" > "${report_dir}/${node}/${pod}.log" &
 
     # We don't want to run more than ${max_dump_processes} at a time, so
     # wait once we hit that many nodes. This isn't ideal, since one might
@@ -517,11 +517,8 @@ function dump_nodes_with_logexporter() {
       proc=${max_dump_processes}
       wait
     fi
-  done
   # Wait for any remaining processes.
-  if [[ proc -gt 0 && proc -lt ${max_dump_processes} ]]; then
-    wait
-  fi
+  done; wait)
 
   # List registry of marker files (of nodes whose logexporter succeeded) from GCS.
   local nodes_succeeded