Explainer
Sticky vs rotating proxies, with real session limits
How long a sticky proxy session really lasts at Oxylabs, Decodo, IPRoyal, Bright Data and more, what ends it early, and which mode fits each scraping job.

For every scraping job you need to pick a session mode: a new IP on each request, or one IP held for a while, and if so for how long. Rotate by default, and only stick where cookies or a login have to carry over between steps.
The definitions are easy. What most guides skip is each provider’s real limits, what ends a session early, and the setting that makes a dropped IP fail loudly instead of switching behind your back. That’s what this guide covers, from the vendors’ own docs.
Short on time?
Short on time? Here's my take.
- Scraping many independent pages
- Rotating per request
- Oxylabs and Bright Data rotate by default; Decodo, IPRoyal and Webshare offer a per-request mode. No cookies need to survive between pages.
- A login or multi-step flow on an account you're allowed to use
- Sticky, with the fail-instead flag on
- Set the session a little longer than the flow. Oxylabs, Decodo and Bright Data can answer 502, and IPRoyal 410, instead of switching IP.
- You need the same IP for days
- An ISP or static proxy, not sticky residential
- IPRoyal's 7 days is the longest sticky cap I found; most stop at 24 hours, and the home device behind the IP can drop sooner.
- Collecting search results
- Rotating, or a SERP API instead of proxies
- Each query stands alone, so there's no session to keep.
Prices verified . Sponsors cannot buy a verdict or a position.
What’s the difference between sticky and rotating proxies? #
A rotating proxy sends each request out through a new IP. A sticky session keeps the same exit IP for a window you choose, from seconds up to a day on most plans.
Oxylabs’ residential docs state the default plainly: “These proxies rotate automatically with every request.” Oxylabs. Bright Data’s docs say the same in other words: “By default, Bright Data assigns a random proxy”. Bright Data.
The others offer it as a mode. Decodo’s rotating port “will rotate the IP on every proxy request”, IPRoyal’s Randomize option provides “a new proxy with each request”, and Webshare’s rotate keyword gives “a new IP on every request”. Decodo, IPRoyal, Webshare.
Sticky is the opposite promise, with a catch. Oxylabs says a session “routes multiple requests through the same IP address as long as the IP is online”. Oxylabs session control. On a residential plan that IP belongs to someone’s home device, and it can go away mid-session.
A static or ISP proxy is a third thing: an address you rent per IP for days or months. Webshare’s pricing page bills its static residential product per proxy and says it “suits workflows that need the same IP over a long session”, while rotating residential is “priced per GB of bandwidth”. Webshare pricing.
| Mode | What you get | How long | How it’s billed |
|---|---|---|---|
| Rotating | A new IP on each request | One request | Per GB on the residential plans I read |
| Sticky session | The same IP for a window you set | Seconds to 24 hours on most plans; up to 7 days at IPRoyal | The same per-GB rate |
| Static or ISP | An address you rent | Days to months | Per IP |
Verified 2026-11-17 against the Oxylabs, Decodo, IPRoyal and Webshare pages.
If you’re still choosing the proxy type, my guides to ISP vs residential and residential vs datacenter cover it. This guide stays inside one residential plan.
How do you ask for a sticky IP? #
Two ways: put a session ID in the proxy username (or password), or connect to a sticky port. Change the ID or the port and you get a new IP.
Session IDs are the common method. Every request that carries the same ID goes out through the same IP until the session ends; Bright Data puts it in one line: “Each unique session ID will get a unique IP”. Bright Data config options.
| Provider | Session syntax, as the docs show it | How you set the length |
|---|---|---|
| Oxylabs | sessid-<random alphanumeric> in the username, or a sticky port 10001-49999 |
sesstime in minutes |
| Decodo | user-username-session-example1-sessionduration-90 |
sessionduration in minutes, 1 to 1440, or the dashboard |
| IPRoyal | _session-<8 characters> plus _lifetime-10m |
_lifetime-, one unit, 1 second to 7 days |
| Bright Data | brd-customer-<customer_id>-zone-<zone_name>-session-mystring12345 |
Not documented |
| Webshare | {username}-{country_code}-{geo_params}-{session_id}, with a numeric ID |
Not documented |
| ZenRows | _ttl-30s_session-<id> in the password |
_ttl- values 30s, 1m, 30m, 1h, 24h |
Placeholders as in each vendor’s docs, verified 2026-11-17: Oxylabs, Decodo, IPRoyal, Bright Data, Webshare, ZenRows.
A few formatting rules catch people out. IPRoyal’s session string must be “precisely 8 characters in length”, and for the lifetime “only one unit of time can be specified”. IPRoyal.
Bright Data session IDs can’t contain a -, double quotes or backticks. Bright Data FAQ.
Webshare’s rotate keyword “Cannot be combined with a session ID”, so pick one per request. Webshare. Its separate rotating endpoint, p.webshare.io, is for the proxy-list product and “allows proxies to rotate randomly within your list for each request”. Webshare help.
Sticky ports skip the username change. Oxylabs’ ports 10001-49999 hold an IP “for up to 10 minutes without passing sessid parameters”, and “Reusing the same port keeps the same exit node, while changing the port assigns a new sticky IP”. Oxylabs session control.
The trade-off is targeting: Oxylabs’ sticky ports “do not support city or state-level parameters”, so use a session ID if you need a city. Decodo also lets you pick sticky or rotating from a dashboard drop-down, with sticky lengths of 1, 10, 30 or 60 minutes or a custom value. Decodo.
How long does a sticky session last? #
Ten minutes by default at Oxylabs and Decodo, up to 24 hours at Oxylabs, Decodo and ZenRows, and up to 7 days at IPRoyal. Bright Data and Webshare don’t publish a maximum in the docs I read.
| Provider | Default sticky length | Maximum | Ends after idle | Source |
|---|---|---|---|---|
| Oxylabs | 10 min | 1,440 min (24 h); sticky ports 10 min | 60 s with no requests | Docs |
| Decodo | 10 min | 1,440 min (24 h) | Not stated | Docs |
| IPRoyal | Not stated | 7 days | Not stated | Docs |
| Bright Data | Not stated | Not stated | 5 min | Docs |
| Webshare | Not stated | Not stated | Not stated | Docs |
| ZenRows | Not stated | 24 h | Not stated | Docs |
| SOAX | Unverified | Unverified | Unverified | Session help pages returned 404 |
| NetNut | Unverified | Unverified | Unverified | Docs behind a sign-in |
Residential products only, verified 2026-11-17. “Not stated” means the page I read doesn’t give a number.
Oxylabs’ docs give the clearest rule: “By default, a session lasts 10 minutes, or ends after 60 seconds without any requests – whichever comes first.” The sesstime parameter extends it “beyond the 10-minute default”, “Up to 1440 min (24 hrs)”. Oxylabs session control.
Decodo’s sessionduration “accepts any integer (minute) value between 1 and 1440”, and its default sticky time is 10 minutes. Decodo custom sessions. ZenRows describes its TTL as running “from 30 seconds up to 1 day” but doesn’t name a default. ZenRows.
IPRoyal goes furthest. Its docs set a minimum of 1 second and a maximum of 7 days, and its pricing page says you can “use the same residential proxy as long as you need it - up to 7 days”. IPRoyal docs, IPRoyal pricing. That’s the longest sticky cap I found.
SOAX’s product page says rotation can happen “per request, after N requests, after a set time, or on error”, and that you can “lock requests to a single node with no rotation”, but it gives no duration. SOAX. I couldn’t open its session help articles or NetNut’s docs, so I leave both as unverified rather than repeat numbers from search snippets.
Where vendors contradict their own docs #
Two vendors print a shorter limit on their marketing pages than their docs allow. Decodo’s pricing page says you can keep an IP “for up to 30 min”, while its docs say 1 to 1,440 minutes and a custom duration “up to 24 hours”. Decodo pricing, Decodo docs.
Oxylabs’ blog, last updated July 8, 2025, says “a sticky proxy will hold an IP address for up to 30 minutes”, against 1,440 minutes in its docs. Oxylabs blog. In both cases I go by the docs, because they’re the technical reference and name a parameter you can actually set.
Older guides lag too. Proxyway’s 2022 explainer lists sticky lengths of 1, 10 and 30 minutes; Decodo’s dashboard today offers 1, 10, 30 or 60 minutes plus a custom value. Proxyway, Decodo. The 24-hour caps in my best residential proxies roundup match what I read today.

What ends a session early? #
Five things: the time limit, an idle gap, the home device going offline, a change of targeting parameters, and a switch of region. Only the first is under your full control.
The time limit, even mid-request. Oxylabs is blunt about it: “The session expires once the time limit is reached, even if some requests are still in progress.” Oxylabs session control. Set the length a little longer than the flow, not equal to it.
An idle gap. Oxylabs ends a session after 60 seconds without requests. Bright Data says “Session idle time is 5 minutes”, and after more than 5 minutes idle you get a random proxy even with the same session ID. Oxylabs, Bright Data.
Bright Data’s FAQ suggests a keep-alive request, such as /favicon.ico, “every 30 seconds, to prevent this session from becoming idle for over a minute”. Bright Data FAQ. That works, but it costs bytes, which I price below.
The device goes offline. Decodo warns that on long sessions “the IP will rotate before your specified time due to the residential device at the end going offline”. Oxylabs says “If an exit node drops, sessid will automatically rotate to a new IP”, and IPRoyal says “the system silently assigns a replacement”. Decodo, Oxylabs, IPRoyal.
Proxyway put it well back in 2022: “there’s no guarantee that you’ll keep the same IP for the full duration”. Proxyway. How often that happens, I couldn’t find: no vendor or independent source I read publishes a drop rate.
A parameter changes. Bright Data’s docs say “In case 2 consequent requests carry different country, city or any other parameter: session context will be lost”. It also warns against switching between its global regions (AMER, EMEA, APAC) within a session. Bright Data.
So keep every targeting parameter identical for the life of a session.
Should a dropped IP fail or swap silently? #
Fail. The default at Oxylabs, Decodo and IPRoyal is a silent swap: your next request leaves from a new IP and your flow carries on as if nothing happened. A lock flag makes the provider return an error instead, so you know to restart.
Turn on the fail-instead flag
- Oxylabs: use
sessid_oneip. Then “the system will not rotate to a new IP, and instead the request fails with an HTTP502 Bad Gatewayerror”.- Decodo: use
session_iplock. It “will result with 502 status code and will not grant you a new IP during desired session duration”.- Bright Data: add
-const: “Use the same peer for the session. If peer is unavailable, a 502 error will be returned”.- IPRoyal: add
_killswitch-1, which “returns an HTTP 410 code response instead of rotating”.- Webshare and ZenRows: no equivalent flag in the docs I read.
- SOAX: offers a lock to a single node; what happens when that node drops is unverified.
Flags verified 2026-11-17: Oxylabs, Decodo, Bright Data, IPRoyal, SOAX.
Bright Data has a wrinkle. Its FAQ says that even without -const, the first request after a peer is lost gets “502 - No peers available”, and the next request gets a new peer with the same session ID. Bright Data FAQ. So on Bright Data, treat any 502 inside a session as a signal to restart, flag or no flag.
Why does loud beat silent? Because a silent swap lets you keep spending on a broken flow.
My estimate. Assumptions: a 10-step flow, 500 KB per step, the IP drops at step 8, and you restart from step 1.
Redoing steps 1-7 costs 7 × 500 KB = 3.5 MB whichever way you find out. With a silent swap, if you only notice when step 10 misbehaves, you’ve also spent 3 × 500 KB = 1.5 MB on steps 8-10, for 5 MB in total. With the flag, you stop at step 8 and waste 3.5 MB.
To restart, change the session ID. Bright Data’s FAQ says “To force an IP change, just modify the session ID”, and Decodo says “You can change the port or the session ID to rotate the IP prematurely”. Bright Data FAQ, Decodo.
Which mode for which job? #
Rotate unless state has to carry over between requests. These are my recommendations from the limits above; no vendor sets them.
| Job | Mode | Session length | Why |
|---|---|---|---|
| Independent pages: product pages, price checks | Rotating per request | None | No cookies need to survive between requests |
| Paginating one listing | Rotating if each page URL stands alone; sticky if the site keeps state | A few minutes, kept busy | Idle ends a session after 60 s at Oxylabs and 5 min at Bright Data; keep parameters constant |
| A login or multi-step form on an account you’re allowed to use | Sticky, with the fail flag | The flow’s length plus a margin, under the provider’s cap | Restart the whole flow on a 502 or 410 |
| The same IP for days | ISP or static proxy, billed per IP | Days | Most sticky caps stop at 24 hours, and devices drop |
| Search results | Rotating, or a SERP API instead | None | Each query stands alone |
House recommendations, based on the vendor limits verified 2026-11-17.
For logins, I mean accounts you’re allowed to use, on sites whose terms allow it. Whether the proxies themselves are lawful to use is a separate question, covered in are residential proxies legal.
For search results, a SERP API often makes the proxy question go away, and my guide to scraping APIs vs proxies covers when the trade is worth it. For multi-day IPs, ISP vs residential picks up from here.
One blog claim I don’t build on: ZenRows says “prolonged sticky sessions can increase the chances of anti-bot detection”, but gives no source. ZenRows blog. The reason I’d keep sessions short is simpler: the shorter the session, the less there is to lose when the IP drops.
Do sticky sessions cost more? #
Not per byte on the plans I read. The extra cost comes from bytes you add yourself, through keep-alives and restarted flows.
Decodo’s residential pricing is “bandwidth-based” and lists “Rotating and sticky session options” as included. On its 100 GB plan, that’s 100 × $2.75 = $275 a month either way. Decodo pricing.
IPRoyal charges per GB too, with sticky sessions of up to 7 days on the same plan. On its 10 GB subscription that’s 10 × $5.25 = $52.50. IPRoyal pricing. Both prices verified 2026-11-17.
Keep-alives are where a sticky session quietly adds up.
My estimate. Assumptions: one keep-alive every 30 seconds, as Bright Data’s FAQ suggests; 2 KB per keep-alive round trip; 100 sessions held around the clock; keep-alive bytes billed at the normal rate, which no page I read confirms.
That’s 2 per minute × 60 × 24 = 2,880 keep-alives per session per day, and 2,880 × 2 KB = 5,760 KB, or about 5.8 MB. Across 100 sessions that’s about 0.58 GB a day, or about 17 GB over 30 days. At Decodo’s $2.75 per GB, 17 × $2.75 ≈ $47 a month. I use Decodo’s rate only as an illustration, because I didn’t open Bright Data’s own price for this guide.
So keep sessions alive only while a flow needs them, and let idle ones end. If a job truly needs one IP for days, per-IP pricing is the cheaper shape: Webshare lists static residential at $6 a month for 20 proxies, or $0.30 each, against rotating residential from $3.50 per GB. Webshare pricing. My proxy pricing models guide compares per-GB and per-IP billing properly.

How do you wire sessions into your scraper? #
Give each flow its own proxy session and keep that flow’s cookies with it. When the session dies, throw both away together.
In Playwright, a proxy “can be either set globally for the entire browser, or for each browser context individually”. Playwright. So open one browser context per flow, with that flow’s session ID in the proxy username. Each context keeps its own cookies, so the IP and the cookies travel together.
// A sketch written from Playwright's network docs, not run.
// Put your provider's session syntax in the username (see the table above).
const context = await browser.newContext({
proxy: {
server: 'http://proxy.example.com:8000',
username: 'YOUR_USERNAME_WITH_SESSION_ID',
password: 'YOUR_PASSWORD',
},
});
On a 502 or 410, close that context and open a new one with a fresh session ID, then start the flow again from step 1.
Crawlee builds the same idea in. Its SessionPool handles “rotation of proxy IP addresses along with cookies”, and its examples pass the session ID to newUrl to keep “the same proxyUrl” for a session. Cookies persist per session by default (persistCookiesPerSession is true), and session.retire() throws a session away. Crawlee.
Retire the session on a 502 or 410 and the pool hands you a new one.
What I would do #
I’d leave rotation on as the default for every job. For any flow that carries cookies, I’d set a session ID per flow, a duration a bit longer than the flow, and the provider’s fail-instead flag, then restart the whole flow on a 502 or 410.
I’d keep every targeting parameter fixed for the life of a session, and I wouldn’t send keep-alives unless the flow really pauses longer than the idle limit. If a job needs one IP for more than a day, I’d buy an ISP proxy rather than stretch a residential session.
That’s a recommendation from the docs, not a result from my own checks. If a site starts answering with 429s or blocks, the fix is pace, not new IPs: my guide to web scraping rate limits covers it, and ethical web scraping covers when to stop. The proxies hub maps the rest, and the glossary has the short definition.
FAQ
How long can a sticky session last?
On the docs I read on November 17, 2026: 10 minutes by default at Oxylabs and Decodo, up to 24 hours at Oxylabs, Decodo and ZenRows, and up to 7 days at IPRoyal. Bright Data and Webshare don't publish a maximum. The home device behind the IP can still go offline sooner.
Why did my sticky IP change before the time was up?
Usually one of three things: the residential device behind it went offline, the session sat idle too long (60 seconds at Oxylabs, 5 minutes at Bright Data), or you changed a targeting parameter such as country or city between requests, which ends the session at Bright Data.
Is a sticky session the same as a static or ISP proxy?
No. A sticky session holds a residential IP for minutes to a day, and that IP can drop. A static or ISP proxy is an address you rent per IP for days or months. My ISP vs residential guide covers that choice.
Do sticky sessions cost more?
Not per byte on the plans I read. Decodo and IPRoyal charge the same per-GB rate for sticky and rotating traffic. The extra cost comes from bytes you add yourself: keep-alive requests and flows you have to restart.
How do I force a new IP in a sticky session?
Change the session ID, or the port if you use sticky ports. Bright Data's FAQ says to force an IP change, modify the session ID, and Decodo's docs say changing the port or session ID rotates the IP early.
Are rotating proxies better for avoiding blocks?
ZenRows' blog says prolonged sticky sessions can raise the chance of detection, but it gives no source. If a site is rate-limiting or blocking you, switching IPs isn't the fix: slow down, follow its signals, and stop at a block.
Sources
20 sources, accessed .
Official pricing, docs, changelogs and status pages 17
- Oxylabs: residential proxies docs
- Oxylabs: session control (sessid, sesstime, sticky ports)
- Decodo: residential proxy session types
- Decodo: custom sticky sessions
- Decodo: residential proxy pricing
- IPRoyal: residential proxy rotation
- IPRoyal: residential proxy pricing
- Bright Data: proxy configuration options
- Bright Data: rotating IPs and sessions
- Bright Data: residential proxies FAQ
- Webshare API docs: proxy connection
- Webshare: how to use the rotating proxy endpoint
- Webshare: pricing
- ZenRows: residential proxies sticky TTL
- SOAX: residential proxies
- Playwright: network (proxy per browser context)
- Crawlee: session management


