Random Interview questions Part

The primary difference lies in the Management Plane.

vSS (Standard): Created and managed on each individual ESXi host. If you have 50 hosts, you have 50 individual switches to configure. It is ideal for small environments or isolated host management.
vDS (Distributed): Created at the vCenter Server level and "pushed" down to all associated hosts. It acts as a single virtual switch spanning the entire cluster.
Why use vDS? Beyond centralized management, it unlocks enterprise features like LACP (Link Aggregation), NetFlow (for traffic analysis), and Traffic Shaping. Most importantly, it supports Load Based Teaming (LBT), which dynamically moves traffic to different physical uplinks based on actual congestion, not just a hashing algorithm.

ESXi is unique because it doesn't use a general-purpose OS file system like NTFS or APFS. Instead, it uses:

VMFS (Virtual Machine File System): This is a "clustered" file system. In a traditional file system, if two computers write to the same disk simultaneously, the data gets corrupted. VMFS uses SCSI reservations and ATS (Atomic Test and Set) to allow multiple hosts to access the same storage safely.
vSAN (Virtual SAN): An object-based file system. Instead of traditional LUNs, it pools local storage from across multiple hosts into a single shared datastore.
NFS (Network File System): While not native to the ESXi disk format, ESXi acts as an NFS client to connect to external NAS storage, treating those remote folders as local datastores.

While both are reliable, VMFS 6 is the modern standard for several reasons:

Automatic UNMAP: When you delete a 100GB VM in VMFS 5, that 100GB stays "marked as used" on your physical SAN until you manually run a reclaim command. VMFS 6 does this automatically in the background, keeping your storage thin and efficient.
Block Size: VMFS 6 uses a large 1MB block size but handles small files much better than previous versions through improved metadata management.
Snapshot Support: VMFS 6 supports "SE Sparse" snapshots by default, which are significantly more efficient and perform better during long-term snapshot chains.

When a host shows as "Not Responding," it usually means vCenter has lost its heartbeat connection to the host's management agent (vpxa).

Step 1: Physical/Network Check: Can you ping the Management IP? If not, the issue is likely hardware (power) or the physical switch port.
Step 2: Restart Agents: If you can SSH into the host, run:
- /etc/init.d/hostd restart (The host management service)
- /etc/init.d/vpxa restart (The vCenter agent service)
Step 3: Check Disk Space: Often, the / (root) partition on the ESXi host fills up with logs, causing the management services to crash because they can't write new data.
Step 4: Lockdown Mode: Verify the host hasn't been accidentally placed in "Strict Lockdown Mode," which can block certain management communications.

If a VM refuses to start, the "Why" is usually found in the resource or locking layer:

The Swap File (.vswp): When a VM powers on, ESXi creates a swap file equal to the unreserved RAM. If your datastore is 100% full, the VM cannot create this file and will fail to power on.
File Locks: If a VM was previously running on a host that crashed, the "Lock" on the .vmdk file might still be held by the dead host. You may need to manually break the lock or wait for the HA timeout.
CPU Incompatibility: If you are moving a VM between different types of CPUs (Intel to AMD or different generations), the VM may fail to power on due to missing instruction sets unless EVC (Enhanced vMotion Compatibility) is enabled.

A snapshot is not a backup; it is a temporary point-in-time state.

Workflow: When you click "Take Snapshot," ESXi creates a -delta.vmdk. All new writes go here. The original base disk becomes "Read Only."
The Risk: The longer a snapshot stays open, the larger the delta file grows. This can lead to massive performance degradation (latency) because the host has to read from multiple files to "assemble" one block of data.
Consolidation: This is the process of merging the delta data back into the base disk. If the merge fails (often due to high I/O during the process), you will see a warning: "VM Disk Consolidation Needed." You must run this manually to prevent the VM from crashing due to a full datastore.

RAID 1 (Mirroring): Best for ESXi Boot Drives. It’s simple, survives one disk failure, and has zero "write penalty."
RAID 5 (Parity): Best for General Purpose Storage. It gives you the most usable space (only 1 disk worth of capacity is lost to parity). However, because it has to calculate "parity" for every write, it is slower for write-heavy applications.
RAID 10 (Mirroring + Striping): Best for High-Performance Databases. It offers the speed of RAID 0 and the safety of RAID 1. It is expensive (50% capacity loss) but provides the highest IOPS.

Private IPs are used to ensure your internal data center traffic doesn't collide with the public internet.

Class A (10.0.0.0/8): Huge range. Usually used for the core management backbone of a large enterprise.
Class B (172.16.0.0/12): Medium range. Often used for specific services like vMotion or vSAN traffic.
Class C (192.168.0.0/16): Small range. Typical for home labs, small branch offices, or out-of-band management (iDRAC/iLO).
Significance: Proper IP planning prevents "IP Overlap" when you eventually need to link two data centers via a VPN or private link.

In Linux, everything is organized under the Root (/).

/etc: The most important folder for admins; it contains all configuration files.
/var/log: Where you go when things break.
/dev: Where hardware "lives" as a file (e.g., your hard drive is /dev/sda).
Mounting: In Linux, you don't "switch to the D: drive." Instead, you "mount" the D: drive to a folder (like /mnt/data). This makes the file system look seamless, regardless of how many physical disks are actually behind it.

LearnWithGT