vyralabs

Open infrastructure for Solana validators. We run our node in public.

[vyralabs @vyralabshq]

Vyra Labs is live at https://t.co/1TDssltMr8.

We run a Solana validator in the open and build open-source tooling for the operators who run them.

Everything we learn, we publish. Everything we hit, we fix in code.

Here's the week that made us build it.

[vyralabs @vyralabshq]

@jkdotsol Nice Work🙌🏻

[vyralabs @vyralabshq]

Run our own node in the open.

Ship the tooling that catches what standard monitoring misses.

This is the whole thesis in one thread.

[Jk @jkdotsol]

Three weeks ago I shipped driftwatch and said there was a lot to improve.

Since then it ran on my validator for 3 days. It caught the disk freezing for over a second, 110 times.

The drive's own health check said everything was fine the whole time.

here's everything I found🧵

[vyralabs @vyralabshq]

Then 13.5 hours waiting for 80% of stake to return for the coordinated restart.

Epoch 993 credits: ~60% vs our usual 98-99%.

Full writeup, including the bank hash trap that cost me a wasted snapshot regenerate.
https://t.co/TEyg5K3py4

[vyralabs @vyralabshq]

Testnet stalled on July 21 with a client-level consensus bug.

Our node wedged at slot 423534755. 0.0 slots/sec, ~35k behind.

[vyralabs @vyralabshq]

Switched to stock Agave mid-incident. Five client-specific flags had to come out of the start script or the binary won't boot.

Then the ledger itself was pinned to the pre-stall state, so the fresh client happily resumed a dead fork. Had to move it aside and rebuild.

[vyralabs @vyralabshq]

Testnet went sideways today.

A superminority went delinquent and finalization stalled. The affected nodes were mostly JitoLabs / AgaveBam. Anza advised operators to switch to Agave.

We were running jito 4.2.0-beta.2. Our node wedged, stopped replaying, and ended up ~35k slots behind at 0 slots/sec. Epoch 993 is effectively a write-off for us.

What we did:
- Built stock Agave 4.2.0-beta.2 from source.
- Removed the Jito-specific flags and restarted with `--no-voting`.
- The node was still pinned to a stale ledger slot, so we reset RocksDB and rebuilt from the latest snapshot.
- It's now replaying into the recovered network and catching up.

Not at the tip yet, so we're still not voting. We'll publish the full incident log once recovery is complete.

This wasn't a Jito failure. it was a cluster incident. But it's another reminder that every dependency needs a recovery path.

[vyralabs @vyralabshq]

First full epoch of block production on the Vyra testnet node.

136 of 136 leader slots. Zero skipped. 0.0% skip rate.

Vote success 99% for the epoch, node never fell behind the tip.

220 leader slots assigned for 993.

[Moses @Moses7054]

One more step we took with this one. More to go!

[vyralabs @vyralabshq]

SFDP delegated ~156k SOL of testnet stake to @vyralabshq this week.

Three epochs voting at 98-99% credits got us here.

Epoch 992 assigned us 136 leader slots. First blocks. Skip rate goes public either way.

Live stats: https://t.co/zzrDqDzJjm

[Jk @jkdotsol]

And as always, @SuperteamIN is the cheat code. 🎉

Thanks for believing in us from day one.

Onwards and Upwards 🫡

[Jk @jkdotsol]

@vyralabshq mentioned🎉🙌🏻.

Grateful to @readylayerone and @solana_devs for featuring our work.

[vyralabs @vyralabshq]

Testnet node just cleared every SFDP baseline requirement at epoch 989.

vote credits: 6.8M, needed: 5.6M(85% of cluster avg)
we're voting above the cluster average, not just clearing the bar.

foundation stake is calculated and incoming. when it lands, the before/after goes up here.

[Aseneca💙 @Princeadxisrael]

Another independent validator birthed from the Turbin3 advanced cohorts. We are really proud of what our cadets go on to accomplish within the ecosystem. Well-done to the team and folks who contributed one way or the other to make this a success.

[vyralabs @vyralabshq]

@Princeadxisrael 🙌🏻🫡

[vyralabs @vyralabshq]

@solanaturbine Indeed🙌🏻

[vyralabs @vyralabshq]

Vyra Labs is live at https://t.co/1TDssltMr8.

We run a Solana validator in the open and build open-source tooling for the operators who run them.

Everything we learn, we publish. Everything we hit, we fix in code.

Here's the week that made us build it.

[Moses @Moses7054]

Greatful for all the support!

We are taking one step at a time.

Let's See where we end up!

[Jk @jkdotsol]

Everyone says running a validator is a bad business.

They're probably right.

I didn't start one because I had money to burn.

I started one because I wanted to understand the validator layer and build around it.

That's how @vyralabshq started.

There are still a lot of challenges ahead, and success isn't guaranteed.

But we believe in the @solana ecosystem, the people building it, and the future we're helping build.

[vyralabs @vyralabshq]

@Moses7054 🙌🏻🔥

[bitn8 @qubitn8]

Nothing I like more than seeing new independent validators emerging from @solanaturbine's advanced cohorts. Goated!

[vyralabs @vyralabshq]

The node is live and voting right now.

Bare metal, Singapore, jito-solana v4.2.0-beta.0. XDP on, zero dropped shreds.

Go check it yourself: https://t.co/gd8XSIo7sF

you run a node, follow along.

[vyralabs @vyralabshq]

That's why Vyra Labs exists.

Every production issue becomes open-source infrastructure.

Observability, security, performance, tested on our own node before anyone else runs it.

Check our node performance: https://t.co/okW8EkPPzA