forex-scalping.com

Failover drills sound paranoid until the morning you need them.

Twice a quarter I practise: kill the primary VPS session, confirm the secondary VPS can flatten the same account, and confirm phone app login still works on mobile data. The second broker drill is separate. That one is for account outage, not for mid-trade hedging fantasies. I write pass or fail with dates. A drill I only imagine does not count.

Discretionary scalping dies on delay. Failover is part of the playbook, equal to the entry rule. If the drill fails, I am not live the next London open.

The drill takes about twenty minutes on a Sunday. Things that have failed for me in the past: an expired password on the backup VPS, a phone app logged into the demo server instead of live, and two-factor codes going to an old number. None of those were problems until I tested. Every one of them would have been a disaster with an open position during a freeze.

If you only do one thing, log into your phone app on mobile data and find the close button before Monday.

Expired backup password is the fail that taught me to treat the drill as real. My secondary VPS login had rotated ninety days earlier and I only found out on a Sunday practice kill, with no open trade to save me the panic.

Phone on demo is the other quiet trap. The app remembered the last server, which was the demo from a size check, and the close button on live was one menu deeper than I expected under stress.

Two factor codes going to an old number failed me once as well. I keep a pass or fail log with the date and which leg broke, and a fail means I am not live Monday until it is fixed.

Twenty minutes on Sunday is enough if the checklist is written beforehand. Kill primary, flatten from secondary, confirm phone on mobile data, write the result, done.

Have you ever had a drill pass on the VPS side and fail only on the phone leg, or do both usually break together when something is stale? Phone only fails are the ones I underweighted for years.

LondonScalper, one number I started recording in the drill is the time from killing the primary to being flat on the secondary. My first proper attempt took 3 minutes 40 seconds, mostly waiting for a cold terminal to log in and load history, and with the backup terminal kept logged in all week it is now under 90 seconds.

That brings its own edge case. If the backup terminal also runs the EA, both copies can act on the same signal, so mine keeps algo trading switched off and the drill confirms that setting first.

I also moved the secondary VPS to a different provider in a different city. My original backup sat in the same data centre as the primary, which covers a crashed session but not the outage that took both down for an hour last year.

The cost is latency, around 25 ms to the broker instead of 2 ms. For a machine whose only job is to flatten, that is irrelevant.

Do you time the second broker drill as well, or only check that the login works? Placing a real order there is the part I have never timed.