EX260 Practice Questions
Prepare for EX260 with more than an answer.
- 1
Case Study Scenario:
Company Context:
TechStream Inc. provides high-performance video streaming services. They are migrating from a legacy SAN to a Red Hat Ceph Storage cluster. The cluster consists of 20 nodes, each with 12 NVMe drives. The network is 100GbE.Current Situation:
After the initial deployment, the operations team notices that during high write bursts, the OSDs are consuming excessive memory, leading to occasional OOM (Out of Memory) kills. The currentosd_memory_targetis set to the default value.Requirements:
- Prevent OSD processes from being killed due to memory exhaustion.
- Maintain high write performance for 4K video ingestion.
- Ensure the cluster remains stable during recovery operations.
Question:
Based on the scenario, which tuning adjustment provides the most immediate and effective resolution to the OOM issue while respecting the high-performance hardware?Show answer details
Correct answer: C
The default
osd_memory_target(often 4GB) may be too low for high-performance NVMe drives handling bursty writes, causing the OSD to aggressively cache and potentially spike before it can trim, or conversely, if set too high without calculation, cause OOM. The correct approach is to explicitly calculate the available RAM per OSD (Total RAM - OS Overhead) / Num_OSDs and setosd_memory_targetto this specific value. This informs the BlueStore cache manager exactly how much memory it is allowed to use, preventing the system OOM killer from intervening. - 2
When performing maintenance on a Ceph cluster, you need to restart a specific OSD node. To prevent the cluster from marking the OSDs as 'out' and initiating unnecessary data recovery (rebalancing) during the reboot, which flag should be set on the cluster?
Show answer details
Correct answer: D
The
nooutflag prevents the monitors from marking OSDs as 'out' of the cluster map when they stop reporting. This stops the cluster from automatically starting the rebalancing process (moving data to other OSDs) during temporary outages like a reboot. - 3
A developer wants to access a CephFS file system from a client machine where installing kernel modules is restricted. The client does, however, have the
ceph-commonandceph-fusepackages installed. Which command allows them to mount the file system in user space?Show answer details
Correct answer: A
The
ceph-fusecommand is used to mount the Ceph file system in user space (FUSE), which does not require kernel module privileges or specific kernel version support, making it ideal for restricted environments. - 4
A storage administrator is preparing to deploy a new Red Hat Ceph Storage 5 cluster using
cephadm. The environment consists of 6 nodes: 3 for monitors/managers and 3 for OSDs. The administrator needs to bootstrap the initial cluster on the first node. Which specific command sequence successfully initializes the cluster and prepares it for adding subsequent hosts?Show answer details
Correct answer: A
In Red Hat Ceph Storage 5 and later,
cephadmis the standard deployment tool. Thecephadm bootstrapcommand is used to install the monitor and manager daemons on the first node, generating the necessary keys and configuration files to expand the cluster. - 5
An organization requires an object storage solution that supports high availability across two geographically separated data centers (East and West). The requirement is to have active-active replication where users can write to either zone. Which configuration object must be set up to group these two zones together to share a global namespace?
Show answer details
Correct answer: A
A Zonegroup (formerly known as a region) consists of one or more zones. It defines the geographical boundary for replication. To enable active-active replication where buckets are shared across zones, the zones must be part of the same Zonegroup.
- 6
A system administrator observes that a specific replicated pool 'rbd_prod' is consuming uneven amounts of space across the OSDs. While troubleshooting, they decide to analyze the CRUSH map. Which component of the CRUSH map is primarily responsible for defining the selection logic of how data replicas are placed across failure domains?
Show answer details
Correct answer: D
CRUSH rules define the policy for data placement. They dictate the steps the algorithm takes to select OSDs (e.g., 'take root', 'chooseleaf host', 'emit'). While buckets define the topology, the Rule defines the logic of placement across that topology.
graph TD Rule[CRUSH Rule] -->|Defines Policy| Selection Selection -->|Step 1| Root[Take default] Selection -->|Step 2| Domain[Chooseleaf firstn 3 type host] Selection -->|Output| OSDs[Result: OSD.1, OSD.5, OSD.9]
